* RE: [Bonding-devel] Re: Bonding driver unreliable under high CPU load
@ 2002-09-17 19:28 Cureington, Tony
2002-09-17 19:45 ` Jeff Garzik
0 siblings, 1 reply; 13+ messages in thread
From: Cureington, Tony @ 2002-09-17 19:28 UTC (permalink / raw)
To: Andrew Morton, Pascal Brisset; +Cc: bonding-devel, netdev
I've been running some similar code (on 2.4.18) that makes the ioctl a macro - we must handle the MII ioctls too. The patch below also corrects the mii pointer being assigned the address of a pointer (&ifr.ifr_data, ifr_data is a macro that produces a pointer) instead of the pointer itself.
The patch:
--- linux-2.4.20-pre7/drivers/net/bonding.c Tue Sep 17 09:54:35 2002
+++ linux-2.4.20-pre7_mod/drivers/net/bonding.c Tue Sep 17 11:18:28 2002
@@ -316,6 +316,28 @@
#define IS_UP(dev) ((((dev)->flags & (IFF_UP)) == (IFF_UP)) && \
(netif_running(dev) && netif_carrier_ok(dev)))
+/* this IOCTL macro is used to prevent network drivers from returning -EFAULT
+ * from the ioctl, returning -EFAULT causes a link up status to be returned
+ * from bond_check_dev_link even when the link is even connected. this macro
+ * allows the get_user/copy_from_user in network drivers ioctls to work without
+ * intermittently returning -EFAULT. this turns off argument validity
+ * checking on the address passed to the network driver ioctl.
+ *
+ * this method of turning off argument validity checking is also used in the
+ * following drivers:
+ * /usr/src/linux/drivers/addon/iscsi/; addon/cpip; net/hamradio;
+ * net/wan; sound/;
+ *
+ * ioctl must be set to dev->do_ioctl before this macro
+ */
+#define IOCTL(dev, arg, cmd) ({ \
+ int ret; \
+ mm_segment_t fs = get_fs(); \
+ set_fs(get_ds()); \
+ ret = ioctl(dev, arg, cmd); \
+ set_fs(fs); \
+ ret; })
+
static void bond_restore_slave_flags(slave_t *slave)
{
slave->dev->flags = slave->original_flags;
@@ -416,7 +438,7 @@
/* effect... */
etool.cmd = ETHTOOL_GLINK;
ifr.ifr_data = (char*)&etool;
- if (ioctl(dev, &ifr, SIOCETHTOOL) == 0) {
+ if (IOCTL(dev, &ifr, SIOCETHTOOL) == 0) {
if (etool.data == 1) {
return(MII_LINK_READY);
}
@@ -431,13 +453,13 @@
*/
/* Yes, the mii is overlaid on the ifreq.ifr_ifru */
- mii = (struct mii_ioctl_data *)&ifr.ifr_data;
- if (ioctl(dev, &ifr, SIOCGMIIPHY) != 0) {
+ mii = (struct mii_ioctl_data *)ifr.ifr_data;
+ if (IOCTL(dev, &ifr, SIOCGMIIPHY) != 0) {
return MII_LINK_READY; /* can't tell */
}
mii->reg_num = 1;
- if (ioctl(dev, &ifr, SIOCGMIIREG) == 0) {
+ if (IOCTL(dev, &ifr, SIOCGMIIREG) == 0) {
/*
* mii->val_out contains MII reg 1, BMSR
* 0x0004 means link established
> -----Original Message-----
> From: Andrew Morton [mailto:akpm@digeo.com]
> Sent: Saturday, September 14, 2002 10:24 PM
> To: Pascal Brisset
> Cc: bonding-devel@lists.sourceforge.net; netdev@oss.sgi.com
> Subject: [Bonding-devel] Re: Bonding driver unreliable under high CPU
> load
>
>
> Pascal Brisset wrote:
> >
> > I would like to confirm the problem reported by Tony Cureington at
> >
> http://sourceforge.net/mailarchive/forum.php?thread_id=1015008
> &forum_id=2094
> >
> > Problem: In MII-monitoring mode, when the CPU load is high,
> > the ethernet bonding driver silently fails to detect dead links.
> >
> > How to reproduce:
> > i686, 2.4.19; "modprobe bonding mode=1 miimon=100"; ifenslave two
> > interfaces; ping while you plug/unplug cables. Bonding will
> > switch to the available interface, as expected. Now load the CPU
> > with "while(1) { }", and failover will not work at all anymore.
> >
> > Explanation:
> > The bonding driver monitors the state of its slave interfaces by
> > calling their dev->do_ioctl(SIOCGMIIREG|ETHTOOL_GLINK) from a
> > timer callback function. Whenever this occurs during a user task,
> > the get_user() in the ioctl handling code of the slave fails with
> > -EFAULT because the ifreq struct is allocated in the stack of the
> > timer function, above 0xC0000000. In that case, the bonding driver
> > considers the link up by default.
> >
> > This problem went unnoticed because for most applications, when the
> > active link dies, the host becomes idle and the monitoring function
> > gets a chance to run during a kernel thread (in which case
> it works).
> > The active-backup switchover is just slower than it should be.
> > Serious trouble only happens when the active link dies
> during a long,
> > CPU-intensive job.
> >
> > Is anyone working on a fix ? Maybe running the monitoring stuff in
> > a dedicated task ?
>
> Running the ioctl in interrupt context is bad. Probably what should
> happen here is that the whole link monitoring function be pushed up
> to process context via a schedule_task() callout, or a do it in a
> dedicated kernel thread.
>
> This patch will probably make it work, but the slave device's
> ioctl simply
> isn't designed to be called from this context - it could try to take
> a semaphore, or a non-interrupt-safe lock or anything.
>
> --- linux-2.4.20-pre7/drivers/net/bonding.c Thu Sep 12 20:35:22 2002
> +++ linux-akpm/drivers/net/bonding.c Sat Sep 14 20:23:45 2002
> @@ -208,6 +208,7 @@
> #include <asm/io.h>
> #include <asm/dma.h>
> #include <asm/uaccess.h>
> +#include <asm/processor.h>
> #include <linux/errno.h>
>
> #include <linux/netdevice.h>
> @@ -401,6 +402,7 @@ static u16 bond_check_dev_link(struct ne
> struct ifreq ifr;
> struct mii_ioctl_data *mii;
> struct ethtool_value etool;
> + int ioctl_ret;
>
> if ((ioctl = dev->do_ioctl) != NULL) { /* ioctl to
> access MII */
> /* TODO: set pointer to correct ioctl on a per
> team member */
> @@ -416,7 +418,13 @@ static u16 bond_check_dev_link(struct ne
> /* effect...
> */
> etool.cmd = ETHTOOL_GLINK;
> ifr.ifr_data = (char*)&etool;
> - if (ioctl(dev, &ifr, SIOCETHTOOL) == 0) {
> + {
> + mm_segment_t old_fs = get_fs();
> + set_fs(KERNEL_DS);
> + ioctl_ret = ioctl(dev, &ifr, SIOCETHTOOL);
> + set_fs(old_fs);
> + }
> + if (ioctl_ret == 0) {
> if (etool.data == 1) {
> return(MII_LINK_READY);
> }
>
>
> -
>
>
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf
> _______________________________________________
> Bonding-devel mailing list
> Bonding-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/bonding-devel
>
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [Bonding-devel] Re: Bonding driver unreliable under high CPU load
2002-09-17 19:28 [Bonding-devel] Re: Bonding driver unreliable under high CPU load Cureington, Tony
@ 2002-09-17 19:45 ` Jeff Garzik
2002-09-17 19:53 ` [Bonding-devel] Re: Bonding driver unreliable under high CPUload Andrew Morton
0 siblings, 1 reply; 13+ messages in thread
From: Jeff Garzik @ 2002-09-17 19:45 UTC (permalink / raw)
To: Cureington, Tony; +Cc: Andrew Morton, Pascal Brisset, bonding-devel, netdev
Cureington, Tony wrote:
> /* Yes, the mii is overlaid on the ifreq.ifr_ifru */
> mii = (struct mii_ioctl_data *)&ifr.ifr_data;
> if (ioctl(dev, &ifr, SIOCGMIIPHY) != 0) {
> return MII_LINK_READY; /* can't tell */
> }
>
> mii->reg_num = 1;
> if (ioctl(dev, &ifr, SIOCGMIIREG) == 0) {
> /*
> * mii->val_out contains MII reg 1, BMSR
> * 0x0004 means link established
> */
> return mii->val_out;
> }
Speaking of bonding, I wonder about the above code -- why do you return
mii->val_out directly? AFAICS you should test BMSR_LSTATUS (a.k.a.
0x0004) and return MII_LINK_READY or zero -- not a bunch of random bits.
The status word can certainly be non-zero even when link is absent.
Also, a further question: do you have access to the slave struct
net_device? If so, just test netif_carrier_ok(slave_dev) and avoid all
that ioctl calling if it returns non-zero.
Jeff
^ permalink raw reply [flat|nested] 13+ messages in thread
* RE: [Bonding-devel] Re: Bonding driver unreliable under high CPU load
@ 2002-09-17 19:46 Jay Vosburgh
0 siblings, 0 replies; 13+ messages in thread
From: Jay Vosburgh @ 2002-09-17 19:46 UTC (permalink / raw)
To: Cureington, Tony; +Cc: Andrew Morton, Pascal Brisset, bonding-devel, netdev
Actually, judging from how other drivers do this, the mii_ioctl_data
structure is really supposed to be assigned to &ifr.ifr_data. I dont
believe there is any storage where ifr_data points, and the ifr_ifru union
is 16 bytes, which is the size of the mii_ioctl_data structure.
-J
"Cureington, Tony" <tony.cureington@hp.com>@lists.sourceforge.net on
09/17/2002 12:28:52 PM
Sent by: bonding-devel-admin@lists.sourceforge.net
To: "Andrew Morton" <akpm@digeo.com>, "Pascal Brisset"
<pascal.brisset-ml@wanadoo.fr>
cc: <bonding-devel@lists.sourceforge.net>, <netdev@oss.sgi.com>
Subject: RE: [Bonding-devel] Re: Bonding driver unreliable under high
CPU load
I've been running some similar code (on 2.4.18) that makes the ioctl a
macro - we must handle the MII ioctls too. The patch below also corrects
the mii pointer being assigned the address of a pointer (&ifr.ifr_data,
ifr_data is a macro that produces a pointer) instead of the pointer itself.
The patch:
--- linux-2.4.20-pre7/drivers/net/bonding.c Tue Sep 17 09:54:35 2002
+++ linux-2.4.20-pre7_mod/drivers/net/bonding.c Tue Sep 17 11:18:28 2002
@@ -316,6 +316,28 @@
#define IS_UP(dev) ((((dev)->flags & (IFF_UP)) == (IFF_UP)) && \
(netif_running(dev) && netif_carrier_ok(dev)))
+/* this IOCTL macro is used to prevent network drivers from returning
-EFAULT
+ * from the ioctl, returning -EFAULT causes a link up status to be
returned
+ * from bond_check_dev_link even when the link is even connected. this
macro
+ * allows the get_user/copy_from_user in network drivers ioctls to work
without
+ * intermittently returning -EFAULT. this turns off argument validity
+ * checking on the address passed to the network driver ioctl.
+ *
+ * this method of turning off argument validity checking is also used in
the
+ * following drivers:
+ * /usr/src/linux/drivers/addon/iscsi/; addon/cpip; net/hamradio;
+ * net/wan; sound/;
+ *
+ * ioctl must be set to dev->do_ioctl before this macro
+ */
+#define IOCTL(dev, arg, cmd) ({ \
+ int ret; \
+ mm_segment_t fs = get_fs(); \
+ set_fs(get_ds()); \
+ ret = ioctl(dev, arg, cmd); \
+ set_fs(fs); \
+ ret; })
+
static void bond_restore_slave_flags(slave_t *slave)
{
slave->dev->flags = slave->original_flags;
@@ -416,7 +438,7 @@
/* effect... */
etool.cmd = ETHTOOL_GLINK;
ifr.ifr_data = (char*)&etool;
- if (ioctl(dev, &ifr, SIOCETHTOOL) == 0) {
+ if (IOCTL(dev, &ifr, SIOCETHTOOL) == 0) {
if (etool.data == 1) {
return(MII_LINK_READY);
}
@@ -431,13 +453,13 @@
*/
/* Yes, the mii is overlaid on the ifreq.ifr_ifru */
- mii = (struct mii_ioctl_data *)&ifr.ifr_data;
- if (ioctl(dev, &ifr, SIOCGMIIPHY) != 0) {
+ mii = (struct mii_ioctl_data *)ifr.ifr_data;
+ if (IOCTL(dev, &ifr, SIOCGMIIPHY) != 0) {
return MII_LINK_READY; /* can't tell */
}
mii->reg_num = 1;
- if (ioctl(dev, &ifr, SIOCGMIIREG) == 0) {
+ if (IOCTL(dev, &ifr, SIOCGMIIREG) == 0) {
/*
* mii->val_out contains MII reg 1, BMSR
* 0x0004 means link established
> -----Original Message-----
> From: Andrew Morton [mailto:akpm@digeo.com]
> Sent: Saturday, September 14, 2002 10:24 PM
> To: Pascal Brisset
> Cc: bonding-devel@lists.sourceforge.net; netdev@oss.sgi.com
> Subject: [Bonding-devel] Re: Bonding driver unreliable under high CPU
> load
>
>
> Pascal Brisset wrote:
> >
> > I would like to confirm the problem reported by Tony Cureington at
> >
> http://sourceforge.net/mailarchive/forum.php?thread_id=1015008
> &forum_id=2094
> >
> > Problem: In MII-monitoring mode, when the CPU load is high,
> > the ethernet bonding driver silently fails to detect dead links.
> >
> > How to reproduce:
> > i686, 2.4.19; "modprobe bonding mode=1 miimon=100"; ifenslave two
> > interfaces; ping while you plug/unplug cables. Bonding will
> > switch to the available interface, as expected. Now load the CPU
> > with "while(1) { }", and failover will not work at all anymore.
> >
> > Explanation:
> > The bonding driver monitors the state of its slave interfaces by
> > calling their dev->do_ioctl(SIOCGMIIREG|ETHTOOL_GLINK) from a
> > timer callback function. Whenever this occurs during a user task,
> > the get_user() in the ioctl handling code of the slave fails with
> > -EFAULT because the ifreq struct is allocated in the stack of the
> > timer function, above 0xC0000000. In that case, the bonding driver
> > considers the link up by default.
> >
> > This problem went unnoticed because for most applications, when the
> > active link dies, the host becomes idle and the monitoring function
> > gets a chance to run during a kernel thread (in which case
> it works).
> > The active-backup switchover is just slower than it should be.
> > Serious trouble only happens when the active link dies
> during a long,
> > CPU-intensive job.
> >
> > Is anyone working on a fix ? Maybe running the monitoring stuff in
> > a dedicated task ?
>
> Running the ioctl in interrupt context is bad. Probably what should
> happen here is that the whole link monitoring function be pushed up
> to process context via a schedule_task() callout, or a do it in a
> dedicated kernel thread.
>
> This patch will probably make it work, but the slave device's
> ioctl simply
> isn't designed to be called from this context - it could try to take
> a semaphore, or a non-interrupt-safe lock or anything.
>
> --- linux-2.4.20-pre7/drivers/net/bonding.c Thu Sep 12 20:35:22 2002
> +++ linux-akpm/drivers/net/bonding.c Sat Sep 14 20:23:45 2002
> @@ -208,6 +208,7 @@
> #include <asm/io.h>
> #include <asm/dma.h>
> #include <asm/uaccess.h>
> +#include <asm/processor.h>
> #include <linux/errno.h>
>
> #include <linux/netdevice.h>
> @@ -401,6 +402,7 @@ static u16 bond_check_dev_link(struct ne
> struct ifreq ifr;
> struct mii_ioctl_data *mii;
> struct ethtool_value etool;
> + int ioctl_ret;
>
> if ((ioctl = dev->do_ioctl) != NULL) { /* ioctl to
> access MII */
> /* TODO: set pointer to correct ioctl on a per
> team member */
> @@ -416,7 +418,13 @@ static u16 bond_check_dev_link(struct ne
> /* effect...
> */
> etool.cmd = ETHTOOL_GLINK;
> ifr.ifr_data = (char*)&etool;
> - if (ioctl(dev, &ifr, SIOCETHTOOL) == 0) {
> + {
> + mm_segment_t old_fs = get_fs();
> + set_fs(KERNEL_DS);
> + ioctl_ret = ioctl(dev, &ifr, SIOCETHTOOL);
> + set_fs(old_fs);
> + }
> + if (ioctl_ret == 0) {
> if (etool.data == 1) {
> return(MII_LINK_READY);
> }
>
>
> -
>
>
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf
> _______________________________________________
> Bonding-devel mailing list
> Bonding-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/bonding-devel
>
-------------------------------------------------------
This SF.NET email is sponsored by: AMD - Your access to the experts
on Hammer Technology! Open Source & Linux Developers, register now
for the AMD Developer Symposium. Code: EX8664
http://www.developwithamd.com/developerlab
_______________________________________________
Bonding-devel mailing list
Bonding-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bonding-devel
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [Bonding-devel] Re: Bonding driver unreliable under high CPUload
2002-09-17 19:45 ` Jeff Garzik
@ 2002-09-17 19:53 ` Andrew Morton
2002-09-17 19:58 ` Chad N. Tindel
2002-09-17 20:11 ` Jeff Garzik
0 siblings, 2 replies; 13+ messages in thread
From: Andrew Morton @ 2002-09-17 19:53 UTC (permalink / raw)
To: Jeff Garzik; +Cc: Cureington, Tony, Pascal Brisset, bonding-devel, netdev
Jeff Garzik wrote:
>
> ...
> Also, a further question: do you have access to the slave struct
> net_device? If so, just test netif_carrier_ok(slave_dev) and avoid all
> that ioctl calling if it returns non-zero.
Make that "avoid all that ioctl calling from interrupt context", which
is a bug. Of the box-killing variety ;)
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [Bonding-devel] Re: Bonding driver unreliable under high CPUload
2002-09-17 19:53 ` [Bonding-devel] Re: Bonding driver unreliable under high CPUload Andrew Morton
@ 2002-09-17 19:58 ` Chad N. Tindel
2002-09-17 20:07 ` Jeff Garzik
2002-09-17 20:11 ` Jeff Garzik
1 sibling, 1 reply; 13+ messages in thread
From: Chad N. Tindel @ 2002-09-17 19:58 UTC (permalink / raw)
To: Andrew Morton
Cc: Jeff Garzik, Cureington, Tony, Pascal Brisset, bonding-devel,
netdev
> > Also, a further question: do you have access to the slave struct
> > net_device? If so, just test netif_carrier_ok(slave_dev) and avoid all
> > that ioctl calling if it returns non-zero.
>
> Make that "avoid all that ioctl calling from interrupt context", which
> is a bug. Of the box-killing variety ;)
Will netif_carrier_ok(slave_dev) always work? Do all drivers support the
__LINK_STATE_NOCARRIER flag? If so, is there any reason to call the ioctl
in the first place?
Chad
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [Bonding-devel] Re: Bonding driver unreliable under high CPU load
@ 2002-09-17 20:01 Jay Vosburgh
2002-09-17 20:15 ` Jeff Garzik
0 siblings, 1 reply; 13+ messages in thread
From: Jay Vosburgh @ 2002-09-17 20:01 UTC (permalink / raw)
To: Jeff Garzik
Cc: Cureington, Tony, Andrew Morton, Pascal Brisset, bonding-devel,
netdev
Well, now that it's been pointed out to me, that does look pretty
grotty. It works because MII_LINK_READY is defined to be 4, and the return
from bond_check_dev_link() is always a bitwise test against MII_LINK_READY,
so it works. Could be cleaner, though.
As far as netif_carrier_ok() goes, is it reliable? In looking at the
drivers, it appears that some don't update the flag (e.g., 3c59x.c).
-J
Jeff Garzik <jgarzik@mandrakesoft.com>@lists.sourceforge.net on 09/17/2002
12:45:57 PM
Sent by: bonding-devel-admin@lists.sourceforge.net
To: "Cureington, Tony" <tony.cureington@hp.com>
cc: Andrew Morton <akpm@digeo.com>, Pascal Brisset
<pascal.brisset-ml@wanadoo.fr>, bonding-devel@lists.sourceforge.net,
netdev@oss.sgi.com
Subject: Re: [Bonding-devel] Re: Bonding driver unreliable under high
CPU load
Cureington, Tony wrote:
> /* Yes, the mii is overlaid on the ifreq.ifr_ifru */
> mii = (struct mii_ioctl_data *)&ifr.ifr_data;
> if (ioctl(dev, &ifr, SIOCGMIIPHY) != 0) {
> return MII_LINK_READY; /* can't tell */
> }
>
> mii->reg_num = 1;
> if (ioctl(dev, &ifr, SIOCGMIIREG) == 0) {
> /*
> * mii->val_out contains MII reg 1, BMSR
> * 0x0004 means link established
> */
> return mii->val_out;
> }
Speaking of bonding, I wonder about the above code -- why do you return
mii->val_out directly? AFAICS you should test BMSR_LSTATUS (a.k.a.
0x0004) and return MII_LINK_READY or zero -- not a bunch of random bits.
The status word can certainly be non-zero even when link is absent.
Also, a further question: do you have access to the slave struct
net_device? If so, just test netif_carrier_ok(slave_dev) and avoid all
that ioctl calling if it returns non-zero.
Jeff
-------------------------------------------------------
This SF.NET email is sponsored by: AMD - Your access to the experts
on Hammer Technology! Open Source & Linux Developers, register now
for the AMD Developer Symposium. Code: EX8664
http://www.developwithamd.com/developerlab
_______________________________________________
Bonding-devel mailing list
Bonding-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bonding-devel
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [Bonding-devel] Re: Bonding driver unreliable under high CPUload
2002-09-17 19:58 ` Chad N. Tindel
@ 2002-09-17 20:07 ` Jeff Garzik
2002-09-17 20:11 ` Chad N. Tindel
0 siblings, 1 reply; 13+ messages in thread
From: Jeff Garzik @ 2002-09-17 20:07 UTC (permalink / raw)
To: Chad N. Tindel
Cc: Andrew Morton, Cureington, Tony, Pascal Brisset, bonding-devel,
netdev
Chad N. Tindel wrote:
>>>Also, a further question: do you have access to the slave struct
>>>net_device? If so, just test netif_carrier_ok(slave_dev) and avoid all
>>>that ioctl calling if it returns non-zero.
>>
>>Make that "avoid all that ioctl calling from interrupt context", which
>>is a bug. Of the box-killing variety ;)
>
>
> Will netif_carrier_ok(slave_dev) always work? Do all drivers support the
> __LINK_STATE_NOCARRIER flag?
No. Read again the precise language I used :)
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [Bonding-devel] Re: Bonding driver unreliable under high CPUload
2002-09-17 19:53 ` [Bonding-devel] Re: Bonding driver unreliable under high CPUload Andrew Morton
2002-09-17 19:58 ` Chad N. Tindel
@ 2002-09-17 20:11 ` Jeff Garzik
2002-09-17 20:24 ` Andrew Morton
2002-09-17 20:30 ` Andrew Morton
1 sibling, 2 replies; 13+ messages in thread
From: Jeff Garzik @ 2002-09-17 20:11 UTC (permalink / raw)
To: Andrew Morton; +Cc: Cureington, Tony, Pascal Brisset, bonding-devel, netdev
Andrew Morton wrote:
> Jeff Garzik wrote:
>
>>...
>>Also, a further question: do you have access to the slave struct
>>net_device? If so, just test netif_carrier_ok(slave_dev) and avoid all
>>that ioctl calling if it returns non-zero.
>
>
> Make that "avoid all that ioctl calling from interrupt context", which
> is a bug. Of the box-killing variety ;)
Indeed. /me looks at the bond_check_dev_link callers more closely and
shudders.
That wants fixing...
Note that netif_carrier_ok() can indeed be checked in interrupt context.
And if someone wants to send me patches converting more drivers to use
netif_carrier_{on,off}, I would be very happy :)
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [Bonding-devel] Re: Bonding driver unreliable under high CPUload
2002-09-17 20:07 ` Jeff Garzik
@ 2002-09-17 20:11 ` Chad N. Tindel
2002-09-17 20:23 ` Jeff Garzik
0 siblings, 1 reply; 13+ messages in thread
From: Chad N. Tindel @ 2002-09-17 20:11 UTC (permalink / raw)
To: Jeff Garzik
Cc: Andrew Morton, Cureington, Tony, Pascal Brisset, bonding-devel,
netdev
> > Will netif_carrier_ok(slave_dev) always work? Do all drivers support the
> > __LINK_STATE_NOCARRIER flag?
>
>
> No. Read again the precise language I used :)
Right, well thats what I thought. But how can we do what Andrew wants?
Chad
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [Bonding-devel] Re: Bonding driver unreliable under high CPU load
2002-09-17 20:01 Jay Vosburgh
@ 2002-09-17 20:15 ` Jeff Garzik
0 siblings, 0 replies; 13+ messages in thread
From: Jeff Garzik @ 2002-09-17 20:15 UTC (permalink / raw)
To: Jay Vosburgh
Cc: Cureington, Tony, Andrew Morton, Pascal Brisset, bonding-devel,
netdev
Jay Vosburgh wrote:
>
> Well, now that it's been pointed out to me, that does look pretty
> grotty. It works because MII_LINK_READY is defined to be 4, and the return
> from bond_check_dev_link() is always a bitwise test against MII_LINK_READY,
> so it works. Could be cleaner, though.
Yep. Sounds like you also might want to replace a non-standard constant
(MII_LINK_READY) with its standard constant from linux/mii.h,
BMSR_LSTATUS, too, if you are going to use it like this.
> As far as netif_carrier_ok() goes, is it reliable? In looking at the
> drivers, it appears that some don't update the flag (e.g., 3c59x.c).
No. Only some drivers implement it at present -- though all should.
Patches to fix up drivers to use netif_carrier_{on,off} would be very
welcome. There are several examples in-tree to emulate...
Jeff
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [Bonding-devel] Re: Bonding driver unreliable under high CPUload
2002-09-17 20:11 ` Chad N. Tindel
@ 2002-09-17 20:23 ` Jeff Garzik
0 siblings, 0 replies; 13+ messages in thread
From: Jeff Garzik @ 2002-09-17 20:23 UTC (permalink / raw)
To: Chad N. Tindel
Cc: Andrew Morton, Cureington, Tony, Pascal Brisset, bonding-devel,
netdev
Chad N. Tindel wrote:
>>>Will netif_carrier_ok(slave_dev) always work? Do all drivers support the
>>>__LINK_STATE_NOCARRIER flag?
>>
>>
>>No. Read again the precise language I used :)
>
>
> Right, well thats what I thought. But how can we do what Andrew wants?
You'll need execute those ioctls from process context -- or call
dev->do_ioctl() yourself with proper locking. The latter would
certainly be faster but you must be _very_ careful to avoid deadlocks.
The former would certainly be preferred anyway, because calling an ioctl
to check link state means you are hitting the slave_dev's phy, which is
a very expensive operation in and of itself.
Unfortunately I am thinking that locking in bonding.c may need a
re-think if you want to do this sort of stuff :( Semaphores instead of
a spinlocks may be appropriate, depending on the contexts in which
link_state _really_ needs to be checked.
Jeff
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [Bonding-devel] Re: Bonding driver unreliable under high CPUload
2002-09-17 20:11 ` Jeff Garzik
@ 2002-09-17 20:24 ` Andrew Morton
2002-09-17 20:30 ` Andrew Morton
1 sibling, 0 replies; 13+ messages in thread
From: Andrew Morton @ 2002-09-17 20:24 UTC (permalink / raw)
To: Jeff Garzik; +Cc: Cureington, Tony, Pascal Brisset, bonding-devel, netdev
[-- Attachment #1: Type: text/plain, Size: 1799 bytes --]
Jeff Garzik wrote:
>
> Andrew Morton wrote:
> > Jeff Garzik wrote:
> >
> >>...
> >>Also, a further question: do you have access to the slave struct
> >>net_device? If so, just test netif_carrier_ok(slave_dev) and avoid all
> >>that ioctl calling if it returns non-zero.
> >
> >
> > Make that "avoid all that ioctl calling from interrupt context", which
> > is a bug. Of the box-killing variety ;)
>
> Indeed. /me looks at the bond_check_dev_link callers more closely and
> shudders.
>
> That wants fixing...
>
> Note that netif_carrier_ok() can indeed be checked in interrupt context.
> And if someone wants to send me patches converting more drivers to use
> netif_carrier_{on,off}, I would be very happy :)
That would be best.
I'm so slack. I received the below two years ago; Nelson has
added netif_carrier_foo support to 3c59x.c. As Jeff says: patches
solicited.
-------- Original Message --------
Subject: Re: netlink
Date: Wed, 14 Jun 2000 11:19:33 +0800 (SST)
From: Nelson <tanginhw@comp.nus.edu.sg>
To: Andrew Morton <andrewm@uow.edu.au>
References: <394610FF.4052CEAA@uow.edu.au>
hi Andrew,
The modified 3c59x.c is attached.
for your easy reference, these lines in the attached driver
was modified:
69-72: notes on what are added
121-123: defined the time to expiry of the vertex_timer
as TX_EXPIRE. i have set the value as 1*HZ.
the original value should be 60*HZ. you may
set to any value you feel is good ;p you r rite
abt the 1 sec part since reading the MII management
registers is time consuming.
1122: used TX_EXPIRE instead
1335, 1399, 1428:
set next_tick to TX_EXPIRE
1351: added calls to netif_carrier_on()
1356: added calls to netif_carrier_off()
1367-1371: added checking of link state
do let me know if there are any discrepencies.
thanx.
[-- Attachment #2: 3c59x.c --]
[-- Type: text/plain, Size: 84378 bytes --]
/* EtherLinkXL.c: A 3Com EtherLink PCI III/XL ethernet driver for linux. */
/*
Written 1996-1999 by Donald Becker.
This software may be used and distributed according to the terms
of the GNU Public License, incorporated herein by reference.
This driver is for the 3Com "Vortex" and "Boomerang" series ethercards.
Members of the series include Fast EtherLink 3c590/3c592/3c595/3c597
and the EtherLink XL 3c900 and 3c905 cards.
The author may be reached as becker@CESDIS.gsfc.nasa.gov, or C/O
Center of Excellence in Space Data and Information Sciences
Code 930.5, Goddard Space Flight Center, Greenbelt MD 20771
Linux Kernel Additions:
0.99H+lk0.9 - David S. Miller - softnet, PCI DMA updates
0.99H+lk1.0 - Jeff Garzik <jgarzik@mandrakesoft.com>
Remove compatibility defines for kernel versions < 2.2.x.
Update for new 2.3.x module interface
LK1.1.2 (March 19, 2000)
* New PCI interface (jgarzik)
LK1.1.3 25 April 2000, Andrew Morton <andrewm@uow.edu.au>
- Merged with 3c575_cb.c
- Don't set RxComplete in boomerang interrupt enable reg
- spinlock in vortex_timer to protect mdio functions
- disable local interrupts around call to vortex_interrupt in
vortex_tx_timeout() (So vortex_interrupt can use spin_lock())
- Select window 3 in vortex_timer()'s write to Wn3_MAC_Ctrl
- In vortex_start_xmit(), move the lock to _after_ we've altered
vp->cur_tx and vp->tx_full. This defeats the race between
vortex_start_xmit() and vortex_interrupt which was identified
by Bogdan Costescu.
- Merged back support for six new cards from various sources
- Set vortex_have_pci if pci_module_init returns zero (fixes cardbus
insertion oops)
- Tell it that 3c905C has NWAY for 100bT autoneg
- Fix handling of SetStatusEnd in 'Too much work..' code, as
per 2.3.99's 3c575_cb (Dave Hinds).
- Split ISR into two for vortex & boomerang
- Fix MOD_INC/DEC races
- Handle resource allocation failures.
- Fix 3CCFE575CT LED polarity
- Make tx_interrupt_mitigation the default
LK1.1.4 25 April 2000, Andrew Morton <andrewm@uow.edu.au>
- Add extra TxReset to vortex_up() to fix 575_cb hotplug initialisation
probs.
- Put vortex_info_tbl into __devinitdata
- In the vortex_error StatsFull HACK, disable stats in vp->intr_enable
as well as in the hardware.
- Increased the loop counter in wait_for_completion from 2,000 to 4,000.
LK1.1.5 28 April 2000, andrewm
- Added powerpc defines
- Some extra diagnostics
- In vortex_error(), reset the Tx on maxCollisions. Otherwise most
chips usually get a Tx timeout.
- Added extra_reset module parm
- Replaced some inline timer manip with mod_timer
(Franois romieu <Francois.Romieu@nic.fr>)
- In vortex_up(), don't make Wn3_config initialisation dependent upon
has_nway (this came across from 3c575_cb).
- See http://www.uow.edu.au/~andrewm/linux/#3c59x-2.3 for more details.
10 June 2000, Nelson Tan Gin Hwa <tanginhw@comp.nus.edu.sg>
- In vortex_timer(), make calls to netif_carrier_on() and
netif_carrier_off() to set the __LINK_STATE_NOCARRIER bit to
reflect the link state status.
*/
/*
* FIXME: This driver _could_ support MTU changing, but doesn't. See Don's
* hamaci.c implementation as well as other drivers
*
* NOTE: If you make 'vortex_debug' a constant (#define vortex_debug 0) the
* driver shrinks by 2k due to dead code elimination. There will be some
* performance benefits from this due to elimination of all the tests and
* reduced cache footprint.
*/
/* "Knobs" that adjust features and parameters. */
/* Set the copy breakpoint for the copy-only-tiny-frames scheme.
Setting to > 1512 effectively disables this feature. */
static const int rx_copybreak = 200;
/* Allow setting MTU to a larger size, bypassing the normal ethernet setup. */
static const int mtu = 1500;
/* Maximum events (Rx packets, etc.) to handle at each interrupt. */
static int max_interrupt_work = 32;
/* Give the NIC an extra reset at the end of vortex_up() */
static int extra_reset = 0;
/* Allow aggregation of Tx interrupts. Saves CPU load at the cost
* of possible Tx stalls if the system is blocking interrupts
* somewhere else. Undefine this to disable.
* AKPM 26 April 2000: enabling this still gets vestigial Tx timeouts
* in a heavily loaded (collision-prone) 10BaseT LAN. Should be OK with
* switched Ethernet.
*/
#define tx_interrupt_mitigation 1
/* Put out somewhat more debugging messages. (0: no msg, 1 minimal .. 6). */
#define vortex_debug debug
#ifdef VORTEX_DEBUG
static int vortex_debug = VORTEX_DEBUG;
#else
static int vortex_debug = 1;
#endif
/* Some values here only for performance evaluation and path-coverage
debugging. */
static int rx_nocopy = 0, rx_copy = 0, queued_packet = 0, rx_csumhits;
/* A few values that may be tweaked. */
/* Time in jiffies before concluding the transmitter is hung. */
#define TX_TIMEOUT ((400*HZ)/1000)
/* Time for the vortex_timer() function to be called again */
/* originally was 60*HZ */
#define TX_EXPIRE 1*HZ
/* Keep the ring sizes a power of two for efficiency. */
#define TX_RING_SIZE 16
#define RX_RING_SIZE 32
#define PKT_BUF_SZ 1536 /* Size of each temporary Rx buffer.*/
#ifndef __OPTIMIZE__
#error You must compile this file with the correct options!
#error See the last lines of the source file.
#error You must compile this driver with "-O".
#endif
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/string.h>
#include <linux/timer.h>
#include <linux/errno.h>
#include <linux/in.h>
#include <linux/ioport.h>
#include <linux/malloc.h>
#include <linux/interrupt.h>
#include <linux/pci.h>
#include <linux/init.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/skbuff.h>
#include <asm/irq.h> /* For NR_IRQS only. */
#include <asm/bitops.h>
#include <asm/io.h>
/* John Daniel <jdaniel@etresoft.com> said these work... */
#ifdef __powerpc__
#define outsl outsl_ns
#define insl insl_ns
#endif
/* Kernel compatibility defines, some common to David Hinds' PCMCIA package.
This is only in the support-all-kernels source code. */
#define RUN_AT(x) (jiffies + (x))
#include <linux/delay.h>
static char version[] __devinitdata =
"3c59x.c:v0.99L+LK1.1.5 30 Apr 2000 Donald Becker and others http://cesdis.gsfc.nasa.gov/linux/drivers/vortex.html " "$Revision: 1.78 $\n";
MODULE_AUTHOR("Donald Becker <becker@cesdis.gsfc.nasa.gov>");
MODULE_DESCRIPTION("3Com 3c59x/3c90x/3c575 series Vortex/Boomerang/Cyclone driver");
MODULE_PARM(debug, "i");
MODULE_PARM(options, "1-" __MODULE_STRING(8) "i");
MODULE_PARM(full_duplex, "1-" __MODULE_STRING(8) "i");
MODULE_PARM(rx_copybreak, "i");
MODULE_PARM(max_interrupt_work, "i");
MODULE_PARM(extra_reset, "i");
MODULE_PARM(compaq_ioaddr, "i");
MODULE_PARM(compaq_irq, "i");
MODULE_PARM(compaq_device_id, "i");
/* Operational parameter that usually are not changed. */
/* The Vortex size is twice that of the original EtherLinkIII series: the
runtime register window, window 1, is now always mapped in.
The Boomerang size is twice as large as the Vortex -- it has additional
bus master control registers. */
#define VORTEX_TOTAL_SIZE 0x20
#define BOOMERANG_TOTAL_SIZE 0x40
/* Set iff a MII transceiver on any interface requires mdio preamble.
This only set with the original DP83840 on older 3c905 boards, so the extra
code size of a per-interface flag is not worthwhile. */
static char mii_preamble_required = 0;
#define PFX "3c59x: "
/*
Theory of Operation
I. Board Compatibility
This device driver is designed for the 3Com FastEtherLink and FastEtherLink
XL, 3Com's PCI to 10/100baseT adapters. It also works with the 10Mbs
versions of the FastEtherLink cards. The supported product IDs are
3c590, 3c592, 3c595, 3c597, 3c900, 3c905
The related ISA 3c515 is supported with a separate driver, 3c515.c, included
with the kernel source or available from
cesdis.gsfc.nasa.gov:/pub/linux/drivers/3c515.html
II. Board-specific settings
PCI bus devices are configured by the system at boot time, so no jumpers
need to be set on the board. The system BIOS should be set to assign the
PCI INTA signal to an otherwise unused system IRQ line.
The EEPROM settings for media type and forced-full-duplex are observed.
The EEPROM media type should be left at the default "autoselect" unless using
10base2 or AUI connections which cannot be reliably detected.
III. Driver operation
The 3c59x series use an interface that's very similar to the previous 3c5x9
series. The primary interface is two programmed-I/O FIFOs, with an
alternate single-contiguous-region bus-master transfer (see next).
The 3c900 "Boomerang" series uses a full-bus-master interface with separate
lists of transmit and receive descriptors, similar to the AMD LANCE/PCnet,
DEC Tulip and Intel Speedo3. The first chip version retains a compatible
programmed-I/O interface that has been removed in 'B' and subsequent board
revisions.
One extension that is advertised in a very large font is that the adapters
are capable of being bus masters. On the Vortex chip this capability was
only for a single contiguous region making it far less useful than the full
bus master capability. There is a significant performance impact of taking
an extra interrupt or polling for the completion of each transfer, as well
as difficulty sharing the single transfer engine between the transmit and
receive threads. Using DMA transfers is a win only with large blocks or
with the flawed versions of the Intel Orion motherboard PCI controller.
The Boomerang chip's full-bus-master interface is useful, and has the
currently-unused advantages over other similar chips that queued transmit
packets may be reordered and receive buffer groups are associated with a
single frame.
With full-bus-master support, this driver uses a "RX_COPYBREAK" scheme.
Rather than a fixed intermediate receive buffer, this scheme allocates
full-sized skbuffs as receive buffers. The value RX_COPYBREAK is used as
the copying breakpoint: it is chosen to trade-off the memory wasted by
passing the full-sized skbuff to the queue layer for all frames vs. the
copying cost of copying a frame to a correctly-sized skbuff.
IIIC. Synchronization
The driver runs as two independent, single-threaded flows of control. One
is the send-packet routine, which enforces single-threaded use by the
dev->tbusy flag. The other thread is the interrupt handler, which is single
threaded by the hardware and other software.
IV. Notes
Thanks to Cameron Spitzer and Terry Murphy of 3Com for providing development
3c590, 3c595, and 3c900 boards.
The name "Vortex" is the internal 3Com project name for the PCI ASIC, and
the EISA version is called "Demon". According to Terry these names come
from rides at the local amusement park.
The new chips support both ethernet (1.5K) and FDDI (4.5K) packet sizes!
This driver only supports ethernet packets because of the skbuff allocation
limit of 4K.
*/
/* This table drives the PCI probe routines. It's mostly boilerplate in all
of the drivers, and will likely be provided by some future kernel.
*/
enum pci_flags_bit {
PCI_USES_IO=1, PCI_USES_MEM=2, PCI_USES_MASTER=4,
PCI_ADDR0=0x10<<0, PCI_ADDR1=0x10<<1, PCI_ADDR2=0x10<<2, PCI_ADDR3=0x10<<3,
};
enum { IS_VORTEX=1, IS_BOOMERANG=2, IS_CYCLONE=4,
EEPROM_230=8, /* AKPM: Uses 0x230 as the base bitmpas for EEPROM reads */
HAS_PWR_CTRL=0x10, HAS_MII=0x20, HAS_NWAY=0x40, HAS_CB_FNS=0x80, };
enum vortex_chips {
CH_3C590 = 0,
CH_3C592,
CH_3C597,
CH_3C595_1,
CH_3C595_2,
CH_3C595_3,
CH_VORTEX,
CH_3C900_1,
CH_3C900_2,
CH_3C900_3,
CH_3C900_4,
CH_3C900_5,
CH_3C900B_FL,
CH_3C905_1,
CH_3C905_2,
CH_3C905B_1,
CH_3C905B_2,
CH_3C905B_FX,
CH_3C905C,
CH_3C980,
CH_3CSOHO100_TX,
CH_3C555,
CH_3C575_1,
CH_3CCFE575,
CH_3CCFE575CT,
CH_3CCFE656,
CH_3CCFEM656,
CH_3C450,
};
/* note: this array directly indexed by above enums, and MUST
* be kept in sync with both the enums above, and the PCI device
* table below
*/
static struct vortex_chip_info {
const char *name;
int flags;
int drv_flags;
int io_size;
} vortex_info_tbl[] __devinitdata = {
{"3c590 Vortex 10Mbps",
PCI_USES_IO|PCI_USES_MASTER, IS_VORTEX, 32, },
{"3c592 EISA 10mbps Demon/Vortex", /* AKPM: from Don's 3c59x_cb.c 0.49H */
PCI_USES_IO|PCI_USES_MASTER, IS_VORTEX, 32, },
{"3c597 EISA Fast Demon/Vortex", /* AKPM: from Don's 3c59x_cb.c 0.49H */
PCI_USES_IO|PCI_USES_MASTER, IS_VORTEX, 32, },
{"3c595 Vortex 100baseTx",
PCI_USES_IO|PCI_USES_MASTER, IS_VORTEX, 32, },
{"3c595 Vortex 100baseT4",
PCI_USES_IO|PCI_USES_MASTER, IS_VORTEX, 32, },
{"3c595 Vortex 100base-MII",
PCI_USES_IO|PCI_USES_MASTER, IS_VORTEX, 32, },
#define EISA_TBL_OFFSET 6 /* Offset of this entry for vortex_eisa_init */
{"3Com Vortex",
PCI_USES_IO|PCI_USES_MASTER, IS_BOOMERANG, 64, },
{"3c900 Boomerang 10baseT",
PCI_USES_IO|PCI_USES_MASTER, IS_BOOMERANG, 64, },
{"3c900 Boomerang 10Mbps Combo",
PCI_USES_IO|PCI_USES_MASTER, IS_BOOMERANG, 64, },
{"3c900 Cyclone 10Mbps TPO", /* AKPM: from Don's 0.99M */
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE, 128, },
{"3c900 Cyclone 10Mbps Combo",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE, 128, },
{"3c900 Cyclone 10Mbps TPC", /* AKPM: from Don's 0.99M */
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE, 128, },
{"3c900B-FL Cyclone 10base-FL",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE, 128, },
{"3c905 Boomerang 100baseTx",
PCI_USES_IO|PCI_USES_MASTER, IS_BOOMERANG|HAS_MII, 64, },
{"3c905 Boomerang 100baseT4",
PCI_USES_IO|PCI_USES_MASTER, IS_BOOMERANG|HAS_MII, 64, },
{"3c905B Cyclone 100baseTx",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE|HAS_NWAY, 128, },
{"3c905B Cyclone 10/100/BNC",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE|HAS_NWAY, 128, },
{"3c905B-FX Cyclone 100baseFx",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE, 128, },
{"3c905C Tornado",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE|HAS_NWAY, 128, },
{"3c980 Cyclone",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE, 128, },
{"3cSOHO100-TX Hurricane",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE, 128, },
{"3c555 Laptop Hurricane",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE, 128, },
{"3c575 Boomerang CardBus",
PCI_USES_IO|PCI_USES_MASTER, IS_BOOMERANG|HAS_MII|EEPROM_230, 64, },
{"3CCFE575 Cyclone CardBus",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE|HAS_NWAY|HAS_CB_FNS|EEPROM_230, 128, },
{"3CCFE575CT Cyclone CardBus",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE|HAS_NWAY|HAS_CB_FNS|EEPROM_230, 128, },
{"3CCFE656 Cyclone CardBus",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE|HAS_NWAY|HAS_CB_FNS|EEPROM_230, 128, },
{"3CCFEM656 Cyclone CardBus",
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE|HAS_NWAY|HAS_CB_FNS|EEPROM_230, 128, },
{"3c450 Cyclone/unknown", /* AKPM: from Don's 0.99N */
PCI_USES_IO|PCI_USES_MASTER, IS_CYCLONE|HAS_NWAY, 128, },
{0,}, /* 0 terminated list. */
};
static struct pci_device_id vortex_pci_tbl[] __devinitdata = {
{ 0x10B7, 0x5900, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C590 },
{ 0x10B7, 0x5920, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C592 },
{ 0x10B7, 0x5970, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C597 },
{ 0x10B7, 0x5950, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C595_1 },
{ 0x10B7, 0x5951, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C595_2 },
{ 0x10B7, 0x5952, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C595_3 },
{ 0x10B7, 0x5900, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_VORTEX },
{ 0x10B7, 0x9000, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C900_1 },
{ 0x10B7, 0x9001, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C900_2 },
{ 0x10B7, 0x9004, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C900_3 },
{ 0x10B7, 0x9005, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C900_4 },
{ 0x10B7, 0x9006, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C900_5 },
{ 0x10B7, 0x900A, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C900B_FL },
{ 0x10B7, 0x9050, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C905_1 },
{ 0x10B7, 0x9051, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C905_2 },
{ 0x10B7, 0x9055, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C905B_1 },
{ 0x10B7, 0x9058, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C905B_2 },
{ 0x10B7, 0x905A, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C905B_FX },
{ 0x10B7, 0x9200, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C905C },
{ 0x10B7, 0x9800, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C980 },
{ 0x10B7, 0x7646, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3CSOHO100_TX },
{ 0x10B7, 0x5055, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C555 },
{ 0x10B7, 0x5057, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C575_1 },
{ 0x10B7, 0x5157, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3CCFE575 },
{ 0x10B7, 0x5257, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3CCFE575CT },
{ 0x10B7, 0x6560, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3CCFE656 },
{ 0x10B7, 0x6562, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3CCFEM656 },
{ 0x10B7, 0x4500, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CH_3C450 },
{0,} /* 0 terminated list. */
};
MODULE_DEVICE_TABLE(pci, vortex_pci_tbl);
/* Operational definitions.
These are not used by other compilation units and thus are not
exported in a ".h" file.
First the windows. There are eight register windows, with the command
and status registers available in each.
*/
#define EL3WINDOW(win_num) outw(SelectWindow + (win_num), ioaddr + EL3_CMD)
#define EL3_CMD 0x0e
#define EL3_STATUS 0x0e
/* The top five bits written to EL3_CMD are a command, the lower
11 bits are the parameter, if applicable.
Note that 11 parameters bits was fine for ethernet, but the new chip
can handle FDDI length frames (~4500 octets) and now parameters count
32-bit 'Dwords' rather than octets. */
enum vortex_cmd {
TotalReset = 0<<11, SelectWindow = 1<<11, StartCoax = 2<<11,
RxDisable = 3<<11, RxEnable = 4<<11, RxReset = 5<<11,
UpStall = 6<<11, UpUnstall = (6<<11)+1,
DownStall = (6<<11)+2, DownUnstall = (6<<11)+3,
RxDiscard = 8<<11, TxEnable = 9<<11, TxDisable = 10<<11, TxReset = 11<<11,
FakeIntr = 12<<11, AckIntr = 13<<11, SetIntrEnb = 14<<11,
SetStatusEnb = 15<<11, SetRxFilter = 16<<11, SetRxThreshold = 17<<11,
SetTxThreshold = 18<<11, SetTxStart = 19<<11,
StartDMAUp = 20<<11, StartDMADown = (20<<11)+1, StatsEnable = 21<<11,
StatsDisable = 22<<11, StopCoax = 23<<11, SetFilterBit = 25<<11,};
/* The SetRxFilter command accepts the following classes: */
enum RxFilter {
RxStation = 1, RxMulticast = 2, RxBroadcast = 4, RxProm = 8 };
/* Bits in the general status register. */
enum vortex_status {
IntLatch = 0x0001, HostError = 0x0002, TxComplete = 0x0004,
TxAvailable = 0x0008, RxComplete = 0x0010, RxEarly = 0x0020,
IntReq = 0x0040, StatsFull = 0x0080,
DMADone = 1<<8, DownComplete = 1<<9, UpComplete = 1<<10,
DMAInProgress = 1<<11, /* DMA controller is still busy.*/
CmdInProgress = 1<<12, /* EL3_CMD is still busy.*/
};
/* Register window 1 offsets, the window used in normal operation.
On the Vortex this window is always mapped at offsets 0x10-0x1f. */
enum Window1 {
TX_FIFO = 0x10, RX_FIFO = 0x10, RxErrors = 0x14,
RxStatus = 0x18, Timer=0x1A, TxStatus = 0x1B,
TxFree = 0x1C, /* Remaining free bytes in Tx buffer. */
};
enum Window0 {
Wn0EepromCmd = 10, /* Window 0: EEPROM command register. */
Wn0EepromData = 12, /* Window 0: EEPROM results register. */
IntrStatus=0x0E, /* Valid in all windows. */
};
enum Win0_EEPROM_bits {
EEPROM_Read = 0x80, EEPROM_WRITE = 0x40, EEPROM_ERASE = 0xC0,
EEPROM_EWENB = 0x30, /* Enable erasing/writing for 10 msec. */
EEPROM_EWDIS = 0x00, /* Disable EWENB before 10 msec timeout. */
};
/* EEPROM locations. */
enum eeprom_offset {
PhysAddr01=0, PhysAddr23=1, PhysAddr45=2, ModelID=3,
EtherLink3ID=7, IFXcvrIO=8, IRQLine=9,
NodeAddr01=10, NodeAddr23=11, NodeAddr45=12,
DriverTune=13, Checksum=15};
enum Window2 { /* Window 2. */
Wn2_ResetOptions=12,
};
enum Window3 { /* Window 3: MAC/config bits. */
Wn3_Config=0, Wn3_MAC_Ctrl=6, Wn3_Options=8,
};
union wn3_config {
int i;
struct w3_config_fields {
unsigned int ram_size:3, ram_width:1, ram_speed:2, rom_size:2;
int pad8:8;
unsigned int ram_split:2, pad18:2, xcvr:4, autoselect:1;
int pad24:7;
} u;
};
enum Window4 { /* Window 4: Xcvr/media bits. */
Wn4_FIFODiag = 4, Wn4_NetDiag = 6, Wn4_PhysicalMgmt=8, Wn4_Media = 10,
};
enum Win4_Media_bits {
Media_SQE = 0x0008, /* Enable SQE error counting for AUI. */
Media_10TP = 0x00C0, /* Enable link beat and jabber for 10baseT. */
Media_Lnk = 0x0080, /* Enable just link beat for 100TX/100FX. */
Media_LnkBeat = 0x0800,
};
enum Window7 { /* Window 7: Bus Master control. */
Wn7_MasterAddr = 0, Wn7_MasterLen = 6, Wn7_MasterStatus = 12,
};
/* Boomerang bus master control registers. */
enum MasterCtrl {
PktStatus = 0x20, DownListPtr = 0x24, FragAddr = 0x28, FragLen = 0x2c,
TxFreeThreshold = 0x2f, UpPktStatus = 0x30, UpListPtr = 0x38,
};
/* The Rx and Tx descriptor lists.
Caution Alpha hackers: these types are 32 bits! Note also the 8 byte
alignment contraint on tx_ring[] and rx_ring[]. */
#define LAST_FRAG 0x80000000 /* Last Addr/Len pair in descriptor. */
struct boom_rx_desc {
u32 next; /* Last entry points to 0. */
s32 status;
u32 addr; /* Up to 63 addr/len pairs possible. */
s32 length; /* Set LAST_FRAG to indicate last pair. */
};
/* Values for the Rx status entry. */
enum rx_desc_status {
RxDComplete=0x00008000, RxDError=0x4000,
/* See boomerang_rx() for actual error bits */
IPChksumErr=1<<25, TCPChksumErr=1<<26, UDPChksumErr=1<<27,
IPChksumValid=1<<29, TCPChksumValid=1<<30, UDPChksumValid=1<<31,
};
struct boom_tx_desc {
u32 next; /* Last entry points to 0. */
s32 status; /* bits 0:12 length, others see below. */
u32 addr;
s32 length;
};
/* Values for the Tx status entry. */
enum tx_desc_status {
CRCDisable=0x2000, TxDComplete=0x8000,
AddIPChksum=0x02000000, AddTCPChksum=0x04000000, AddUDPChksum=0x08000000,
TxIntrUploaded=0x80000000, /* IRQ when in FIFO, but maybe not sent. */
};
/* Chip features we care about in vp->capabilities, read from the EEPROM. */
enum ChipCaps { CapBusMaster=0x20, CapPwrMgmt=0x2000 };
struct vortex_private {
/* The Rx and Tx rings should be quad-word-aligned. */
struct boom_rx_desc* rx_ring;
struct boom_tx_desc* tx_ring;
dma_addr_t rx_ring_dma;
dma_addr_t tx_ring_dma;
/* The addresses of transmit- and receive-in-place skbuffs. */
struct sk_buff* rx_skbuff[RX_RING_SIZE];
struct sk_buff* tx_skbuff[TX_RING_SIZE];
struct net_device *next_module; /* NULL if PCI device */
unsigned int cur_rx, cur_tx; /* The next free ring entry */
unsigned int dirty_rx, dirty_tx; /* The ring entries to be free()ed. */
struct net_device_stats stats;
struct sk_buff *tx_skb; /* Packet being eaten by bus master ctrl. */
dma_addr_t tx_skb_dma; /* Allocated DMA address for bus master ctrl DMA. */
/* PCI configuration space information. */
struct pci_dev *pdev;
char *cb_fn_base; /* CardBus function status addr space. */
/* The remainder are related to chip state, mostly media selection. */
struct timer_list timer; /* Media selection timer. */
int options; /* User-settable misc. driver options. */
unsigned int media_override:4, /* Passed-in media type. */
default_media:4, /* Read from the EEPROM/Wn3_Config. */
full_duplex:1, force_fd:1, autoselect:1,
bus_master:1, /* Vortex can only do a fragment bus-m. */
full_bus_master_tx:1, full_bus_master_rx:2, /* Boomerang */
hw_csums:1, /* Has hardware checksums. */
tx_full:1,
has_nway:1,
open:1;
int tx_reset_resume; /* Flag to retart timer after vortex_error handling */
u16 status_enable;
u16 intr_enable;
u16 available_media; /* From Wn3_Options. */
u16 capabilities, info1, info2; /* Various, from EEPROM. */
u16 advertising; /* NWay media advertisement */
unsigned char phys[2]; /* MII device addresses. */
u16 deferred; /* Resend these interrupts when we
* bale from the ISR */
u16 io_size; /* Size of PCI region (for release_region) */
spinlock_t lock;
};
/* The action to take with a media selection timer tick.
Note that we deviate from the 3Com order by checking 10base2 before AUI.
*/
enum xcvr_types {
XCVR_10baseT=0, XCVR_AUI, XCVR_10baseTOnly, XCVR_10base2, XCVR_100baseTx,
XCVR_100baseFx, XCVR_MII=6, XCVR_NWAY=8, XCVR_ExtMII=9, XCVR_Default=10,
};
static struct media_table {
char *name;
unsigned int media_bits:16, /* Bits to set in Wn4_Media register. */
mask:8, /* The transceiver-present bit in Wn3_Config.*/
next:8; /* The media type to try next. */
int wait; /* Time before we check media status. */
} media_tbl[] = {
{ "10baseT", Media_10TP,0x08, XCVR_10base2, (14*HZ)/10},
{ "10Mbs AUI", Media_SQE, 0x20, XCVR_Default, (1*HZ)/10},
{ "undefined", 0, 0x80, XCVR_10baseT, 10000},
{ "10base2", 0, 0x10, XCVR_AUI, (1*HZ)/10},
{ "100baseTX", Media_Lnk, 0x02, XCVR_100baseFx, (14*HZ)/10},
{ "100baseFX", Media_Lnk, 0x04, XCVR_MII, (14*HZ)/10},
{ "MII", 0, 0x41, XCVR_10baseT, 3*HZ },
{ "undefined", 0, 0x01, XCVR_10baseT, 10000},
{ "Autonegotiate", 0, 0x41, XCVR_10baseT, 3*HZ},
{ "MII-External", 0, 0x41, XCVR_10baseT, 3*HZ },
{ "Default", 0, 0xFF, XCVR_10baseT, 10000},
};
static int vortex_probe1(struct pci_dev *pdev, long ioaddr, int irq,
int chip_idx, int card_idx);
static void vortex_up(struct net_device *dev);
static void vortex_down(struct net_device *dev);
static int vortex_open(struct net_device *dev);
static void mdio_sync(long ioaddr, int bits);
static int mdio_read(long ioaddr, int phy_id, int location);
static void mdio_write(long ioaddr, int phy_id, int location, int value);
static void vortex_timer(unsigned long arg);
static int vortex_start_xmit(struct sk_buff *skb, struct net_device *dev);
static int boomerang_start_xmit(struct sk_buff *skb, struct net_device *dev);
static int vortex_rx(struct net_device *dev);
static int boomerang_rx(struct net_device *dev);
static void vortex_interrupt(int irq, void *dev_id, struct pt_regs *regs);
static void boomerang_interrupt(int irq, void *dev_id, struct pt_regs *regs);
static int vortex_close(struct net_device *dev);
static void dump_tx_ring(struct net_device *dev);
static void update_stats(long ioaddr, struct net_device *dev);
static struct net_device_stats *vortex_get_stats(struct net_device *dev);
static void set_rx_mode(struct net_device *dev);
static int vortex_ioctl(struct net_device *dev, struct ifreq *rq, int cmd);
static void vortex_tx_timeout(struct net_device *dev);
static void acpi_set_WOL(struct net_device *dev);
\f
/* This driver uses 'options' to pass the media type, full-duplex flag, etc. */
/* Option count limit only -- unlimited interfaces are supported. */
#define MAX_UNITS 8
static int options[MAX_UNITS] = { -1, -1, -1, -1, -1, -1, -1, -1,};
static int full_duplex[MAX_UNITS] = {-1, -1, -1, -1, -1, -1, -1, -1};
/* A list of all installed Vortex EISA devices, for removing the driver module. */
static struct net_device *root_vortex_eisa_dev = NULL;
/* Variables to work-around the Compaq PCI BIOS32 problem. */
static int compaq_ioaddr = 0, compaq_irq = 0, compaq_device_id = 0x5900;
static int vortex_cards_found = 0;
static void vortex_suspend (struct pci_dev *pdev)
{
struct net_device *dev = pdev->driver_data;
printk(KERN_DEBUG "vortex_suspend(%s)\n", dev->name);
if (dev && dev->priv) {
struct vortex_private *vp = (struct vortex_private *)dev->priv;
if (vp->open) {
netif_device_detach(dev);
vortex_down(dev);
}
}
}
static void vortex_resume (struct pci_dev *pdev)
{
struct net_device *dev = pdev->driver_data;
printk(KERN_DEBUG "vortex_resume(%s)\n", dev->name);
if (dev && dev->priv) {
struct vortex_private *vp = (struct vortex_private *)dev->priv;
if (vp->open) {
vortex_up(dev);
netif_device_attach(dev);
}
}
}
/* returns count found (>= 0), or negative on error */
static int __init vortex_eisa_init (void)
{
long ioaddr;
int rc;
int orig_cards_found = vortex_cards_found;
/* Now check all slots of the EISA bus. */
if (!EISA_bus)
return 0;
for (ioaddr = 0x1000; ioaddr < 0x9000; ioaddr += 0x1000) {
int device_id;
if (request_region(ioaddr, VORTEX_TOTAL_SIZE, "vortex") == NULL)
continue;
/* Check the standard EISA ID register for an encoded '3Com'. */
if (inw(ioaddr + 0xC80) != 0x6d50) {
release_region (ioaddr, VORTEX_TOTAL_SIZE);
continue;
}
/* Check for a product that we support, 3c59{2,7} any rev. */
device_id = (inb(ioaddr + 0xC82)<<8) + inb(ioaddr + 0xC83);
if ((device_id & 0xFF00) != 0x5900) {
release_region (ioaddr, VORTEX_TOTAL_SIZE);
continue;
}
rc = vortex_probe1(NULL, ioaddr, inw(ioaddr + 0xC88) >> 12,
EISA_TBL_OFFSET,
vortex_cards_found);
if (rc == 0)
vortex_cards_found++;
else
release_region (ioaddr, VORTEX_TOTAL_SIZE);
}
/* Special code to work-around the Compaq PCI BIOS32 problem. */
if (compaq_ioaddr) {
vortex_probe1(NULL, compaq_ioaddr, compaq_irq,
compaq_device_id, vortex_cards_found++);
}
return vortex_cards_found - orig_cards_found;
}
/* returns count (>= 0), or negative on error */
static int __devinit vortex_init_one (struct pci_dev *pdev,
const struct pci_device_id *ent)
{
int rc;
rc = vortex_probe1 (pdev, pci_resource_start (pdev, 0), pdev->irq,
ent->driver_data, vortex_cards_found);
if (rc == 0)
vortex_cards_found++;
return rc;
}
/*
* Start up the PCI device which is described by *pdev.
* Return 0 on success.
*
* NOTE: pdev can be NULL, for the case of an EISA driver
*/
static int __devinit vortex_probe1(struct pci_dev *pdev,
long ioaddr, int irq,
int chip_idx, int card_idx)
{
struct vortex_private *vp;
int option;
unsigned int eeprom[0x40], checksum = 0; /* EEPROM contents */
int i;
struct net_device *dev;
static int printed_version = 0;
int retval;
if (!printed_version) {
printk (KERN_INFO "%s", version);
printed_version = 1;
}
dev = init_etherdev(NULL, sizeof(*vp));
if (!dev) {
printk (KERN_ERR PFX "unable to allocate etherdev, aborting\n");
retval = -ENOMEM;
goto out;
}
printk(KERN_INFO "%s: 3Com %s %s at 0x%lx, ",
dev->name,
pdev ? "PCI" : "EISA",
vortex_info_tbl[chip_idx].name,
ioaddr);
/* private struct aligned and zeroed by init_etherdev */
vp = dev->priv;
dev->base_addr = ioaddr;
dev->irq = irq;
dev->mtu = mtu;
vp->has_nway = (vortex_info_tbl[chip_idx].drv_flags & HAS_NWAY) ? 1 : 0;
vp->io_size = vortex_info_tbl[chip_idx].io_size;
/* module list only for EISA devices */
if (pdev == NULL) {
vp->next_module = root_vortex_eisa_dev;
root_vortex_eisa_dev = dev;
}
/* PCI-only startup logic */
if (pdev) {
/* EISA resources already marked, so only PCI needs to do this here */
if (!request_region (ioaddr, vortex_info_tbl[chip_idx].io_size,
dev->name)) {
printk (KERN_ERR "%s: Cannot reserve I/O resource 0x%x @ 0x%lx, aborting\n",
dev->name, vortex_info_tbl[chip_idx].io_size, ioaddr);
retval = -EBUSY;
goto free_dev;
}
/* wake up and enable device */
if (pci_enable_device (pdev)) {
retval = -EIO;
goto free_region;
}
/* enable bus-mastering if necessary */
if (vortex_info_tbl[chip_idx].flags & PCI_USES_MASTER)
pci_set_master (pdev);
}
vp->lock = SPIN_LOCK_UNLOCKED;
vp->pdev = pdev;
/* Makes sure rings are at least 16 byte aligned. */
vp->rx_ring = pci_alloc_consistent(pdev, sizeof(struct boom_rx_desc) * RX_RING_SIZE
+ sizeof(struct boom_tx_desc) * TX_RING_SIZE,
&vp->rx_ring_dma);
if (vp->rx_ring == 0)
{
retval = -ENOMEM;
goto free_region;
}
vp->tx_ring = (struct boom_tx_desc *)(vp->rx_ring + RX_RING_SIZE);
vp->tx_ring_dma = vp->rx_ring_dma + sizeof(struct boom_rx_desc) * RX_RING_SIZE;
/* if we are a PCI driver, we store info in pdev->driver_data
* instead of a module list */
if (pdev)
pdev->driver_data = dev;
/* The lower four bits are the media type. */
if (dev->mem_start)
{ /*
* AKPM: ewww.. The 'options' param is passed in as the third arg to the
* LILO 'ether=' argument for non-modular use
*/
option = dev->mem_start;
}
else if (card_idx < MAX_UNITS)
option = options[card_idx];
else
option = -1;
if (option >= 0) {
vp->media_override = ((option & 7) == 2) ? 0 : option & 15;
vp->full_duplex = (option & 0x200) ? 1 : 0;
vp->bus_master = (option & 16) ? 1 : 0;
} else {
vp->media_override = 7;
vp->full_duplex = 0;
vp->bus_master = 0;
}
if (card_idx < MAX_UNITS && full_duplex[card_idx] > 0)
vp->full_duplex = 1;
vp->force_fd = vp->full_duplex;
vp->options = option;
/* Read the station address from the EEPROM. */
EL3WINDOW(0);
{
int base = (vortex_info_tbl[chip_idx].drv_flags & EEPROM_230) ? 0x230 : EEPROM_Read;
for (i = 0; i < 0x40; i++) {
int timer;
outw(base + i, ioaddr + Wn0EepromCmd);
/* Pause for at least 162 us. for the read to take place. */
for (timer = 10; timer >= 0; timer--) {
udelay(162);
if ((inw(ioaddr + Wn0EepromCmd) & 0x8000) == 0)
break;
}
eeprom[i] = inw(ioaddr + Wn0EepromData);
}
}
for (i = 0; i < 0x18; i++)
checksum ^= eeprom[i];
checksum = (checksum ^ (checksum >> 8)) & 0xff;
if (checksum != 0x00) { /* Grrr, needless incompatible change 3Com. */
while (i < 0x21)
checksum ^= eeprom[i++];
checksum = (checksum ^ (checksum >> 8)) & 0xff;
}
if (checksum != 0x00)
printk(" ***INVALID CHECKSUM %4.4x*** ", checksum);
for (i = 0; i < 3; i++)
((u16 *)dev->dev_addr)[i] = htons(eeprom[i + 10]);
for (i = 0; i < 6; i++)
printk("%c%2.2x", i ? ':' : ' ', dev->dev_addr[i]);
EL3WINDOW(2);
for (i = 0; i < 6; i++)
outb(dev->dev_addr[i], ioaddr + i);
#ifdef __sparc__
printk(", IRQ %s\n", __irq_itoa(dev->irq));
#else
printk(", IRQ %d\n", dev->irq);
/* Tell them about an invalid IRQ. */
if (vortex_debug && (dev->irq <= 0 || dev->irq >= NR_IRQS))
printk(KERN_WARNING " *** Warning: IRQ %d is unlikely to work! ***\n",
dev->irq);
#endif
if (pdev && vortex_info_tbl[chip_idx].drv_flags & HAS_CB_FNS) {
u32 fn_st_addr; /* Cardbus function status space */
fn_st_addr = pci_resource_start (pdev, 2);
if (fn_st_addr)
vp->cb_fn_base = ioremap(fn_st_addr, 128);
printk(KERN_INFO "%s: CardBus functions mapped %8.8x->%p\n",
dev->name, fn_st_addr, vp->cb_fn_base);
#if 1 /* AKPM: the 575_cb and 905B LEDs seem OK without this */
if (vortex_pci_tbl[chip_idx].device != 0x5257) {
EL3WINDOW(2);
outw(0x10 | inw(ioaddr + Wn2_ResetOptions), ioaddr + Wn2_ResetOptions);
}
#endif
}
/* Extract our information from the EEPROM data. */
vp->info1 = eeprom[13];
vp->info2 = eeprom[15];
vp->capabilities = eeprom[16];
if (vp->info1 & 0x8000)
{
vp->full_duplex = 1;
printk(KERN_INFO "Full duplex capable\n");
}
{
static const char * ram_split[] = {"5:3", "3:1", "1:1", "3:5"};
union wn3_config config;
EL3WINDOW(3);
vp->available_media = inw(ioaddr + Wn3_Options);
if ((vp->available_media & 0xff) == 0) /* Broken 3c916 */
vp->available_media = 0x40;
config.i = inl(ioaddr + Wn3_Config);
if (vortex_debug > 1)
printk(KERN_DEBUG " Internal config register is %4.4x, "
"transceivers %#x.\n", config.i, inw(ioaddr + Wn3_Options));
printk(KERN_INFO " %dK %s-wide RAM %s Rx:Tx split, %s%s interface.\n",
8 << config.u.ram_size,
config.u.ram_width ? "word" : "byte",
ram_split[config.u.ram_split],
config.u.autoselect ? "autoselect/" : "",
config.u.xcvr > XCVR_ExtMII ? "<invalid transceiver>" :
media_tbl[config.u.xcvr].name);
vp->default_media = config.u.xcvr;
vp->autoselect = config.u.autoselect;
}
if (vp->media_override != 7) {
printk(KERN_INFO " Media override to transceiver type %d (%s).\n",
vp->media_override, media_tbl[vp->media_override].name);
dev->if_port = vp->media_override;
} else
dev->if_port = vp->default_media;
if (dev->if_port == XCVR_MII || dev->if_port == XCVR_NWAY) {
int phy, phy_idx = 0;
EL3WINDOW(4);
mii_preamble_required++;
mii_preamble_required++;
mdio_read(ioaddr, 24, 1);
for (phy = 1; phy <= 32 && phy_idx < sizeof(vp->phys); phy++) {
int mii_status, phyx = phy & 0x1f;
mii_status = mdio_read(ioaddr, phyx, 1);
if (mii_status && mii_status != 0xffff) {
vp->phys[phy_idx++] = phyx;
printk(KERN_INFO " MII transceiver found at address %d,"
" status %4x.\n", phyx, mii_status);
if ((mii_status & 0x0040) == 0)
mii_preamble_required++;
}
}
mii_preamble_required--;
if (phy_idx == 0) {
printk(KERN_WARNING" ***WARNING*** No MII transceivers found!\n");
vp->phys[0] = 24;
} else {
vp->advertising = mdio_read(ioaddr, vp->phys[0], 4);
if (vp->full_duplex) {
/* Only advertise the FD media types. */
vp->advertising &= ~0x02A0;
mdio_write(ioaddr, vp->phys[0], 4, vp->advertising);
}
}
}
if (vp->capabilities & CapPwrMgmt)
acpi_set_WOL(dev);
if (vp->capabilities & CapBusMaster) {
vp->full_bus_master_tx = 1;
printk(KERN_INFO" Enabling bus-master transmits and %s receives.\n",
(vp->info2 & 1) ? "early" : "whole-frame" );
vp->full_bus_master_rx = (vp->info2 & 1) ? 1 : 2;
vp->bus_master = 0; /* AKPM: vortex only */
}
/* The 3c59x-specific entries in the device structure. */
dev->open = &vortex_open;
dev->hard_start_xmit = &vortex_start_xmit;
dev->stop = &vortex_close;
dev->get_stats = &vortex_get_stats;
dev->do_ioctl = &vortex_ioctl;
dev->set_multicast_list = &set_rx_mode;
dev->tx_timeout = &vortex_tx_timeout;
dev->watchdog_timeo = TX_TIMEOUT;
return 0;
free_region:
release_region (ioaddr, vortex_info_tbl[chip_idx].io_size);
free_dev:
unregister_netdev(dev);
kfree (dev);
printk(KERN_ERR PFX "vortex_probe1 fails. Returns %d\n", retval);
out:
return retval;
}
static void wait_for_completion(struct net_device *dev, int cmd)
{
int i = 4000;
outw(cmd, dev->base_addr + EL3_CMD);
while (--i > 0)
{
if (!(inw(dev->base_addr + EL3_STATUS) & CmdInProgress))
return;
}
printk(KERN_ERR "%s: command 0x%04x did not complete! Status=0x%x\n",
dev->name, cmd, inw(dev->base_addr + EL3_STATUS));
}
static void
vortex_up(struct net_device *dev)
{
long ioaddr = dev->base_addr;
struct vortex_private *vp = (struct vortex_private *)dev->priv;
union wn3_config config;
int i, device_id;
if (vp->pdev)
device_id = vp->pdev->device;
else
device_id = 0x5900; /* EISA */
/* Before initializing select the active media port. */
EL3WINDOW(3);
config.i = inl(ioaddr + Wn3_Config);
if (vp->media_override != 7) {
if (vortex_debug > 1)
printk(KERN_INFO "%s: Media override to transceiver %d (%s).\n",
dev->name, vp->media_override,
media_tbl[vp->media_override].name);
dev->if_port = vp->media_override;
} else if (vp->autoselect) {
if (vp->has_nway) {
printk(KERN_INFO "%s: using NWAY autonegotiation\n", dev->name);
dev->if_port = XCVR_NWAY;
} else {
/* Find first available media type, starting with 100baseTx. */
dev->if_port = XCVR_100baseTx;
while (! (vp->available_media & media_tbl[dev->if_port].mask))
dev->if_port = media_tbl[dev->if_port].next;
printk(KERN_INFO "%s: first available media type: %s\n",
dev->name,
media_tbl[dev->if_port].name);
}
} else {
dev->if_port = vp->default_media;
printk(KERN_INFO "%s: using default media %s\n",
dev->name, media_tbl[dev->if_port].name);
}
init_timer(&vp->timer);
vp->timer.expires = RUN_AT(TX_EXPIRE);
// vp->timer.expires = RUN_AT(media_tbl[dev->if_port].wait);
vp->timer.data = (unsigned long)dev;
vp->timer.function = &vortex_timer; /* timer handler */
add_timer(&vp->timer);
if (vortex_debug > 1)
printk(KERN_DEBUG "%s: Initial media type %s.\n",
dev->name, media_tbl[dev->if_port].name);
vp->full_duplex = vp->force_fd;
config.u.xcvr = dev->if_port;
//AKPM if (!vp->has_nway)
{
if (vortex_debug > 6)
printk(KERN_DEBUG "vortex_up(): writing 0x%x to InternalConfig\n",
config.i);
outl(config.i, ioaddr + Wn3_Config);
}
if (dev->if_port == XCVR_MII || dev->if_port == XCVR_NWAY) {
int mii_reg1, mii_reg5;
EL3WINDOW(4);
/* Read BMSR (reg1) only to clear old status. */
mii_reg1 = mdio_read(ioaddr, vp->phys[0], 1);
mii_reg5 = mdio_read(ioaddr, vp->phys[0], 5);
if (mii_reg5 == 0xffff || mii_reg5 == 0x0000)
netif_carrier_off(dev);/* No MII device or no link partner report */
else {
netif_carrier_on(dev);
if ((mii_reg5 & 0x0100) != 0 /* 100baseTx-FD */
|| (mii_reg5 & 0x00C0) == 0x0040) /* 10T-FD, but not 100-HD */
vp->full_duplex = 1;
}
if (vortex_debug > 1)
printk(KERN_INFO "%s: MII #%d status %4.4x, link partner capability %4.4x,"
" setting %s-duplex.\n", dev->name, vp->phys[0],
mii_reg1, mii_reg5, vp->full_duplex ? "full" : "half");
EL3WINDOW(3);
}
/* Set the full-duplex bit. */
outb(((vp->info1 & 0x8000) || vp->full_duplex ? 0x20 : 0) |
(dev->mtu > 1500 ? 0x40 : 0), ioaddr + Wn3_MAC_Ctrl);
if (vortex_debug > 1) {
printk(KERN_DEBUG "%s: vortex_up() InternalConfig %8.8x.\n",
dev->name, config.i);
}
wait_for_completion(dev, TxReset);
wait_for_completion(dev, RxReset);
outw(SetStatusEnb | 0x00, ioaddr + EL3_CMD);
if (vortex_debug > 1) {
EL3WINDOW(4);
printk(KERN_DEBUG "%s: vortex_up() irq %d media status %4.4x.\n",
dev->name, dev->irq, inw(ioaddr + Wn4_Media));
}
/* Set the station address and mask in window 2 each time opened. */
EL3WINDOW(2);
for (i = 0; i < 6; i++)
outb(dev->dev_addr[i], ioaddr + i);
for (; i < 12; i+=2)
outw(0, ioaddr + i);
if (vp->cb_fn_base) {
u_short n = inw(ioaddr + Wn2_ResetOptions);
#if 0 /* AKPM: This is done in vortex_probe1, and seems to be wrong anyway... */
/* Inverted LED polarity */
if (device_id != 0x5257)
n |= 0x0010;
#endif
/* Inverted polarity of MII power bit */
if ((device_id == 0x6560) || (device_id == 0x6562) ||
(device_id == 0x5257))
n |= 0x4000;
outw(n, ioaddr + Wn2_ResetOptions);
}
if (dev->if_port == XCVR_10base2)
/* Start the thinnet transceiver. We should really wait 50ms...*/
outw(StartCoax, ioaddr + EL3_CMD);
if (dev->if_port != XCVR_NWAY) {
EL3WINDOW(4);
outw((inw(ioaddr + Wn4_Media) & ~(Media_10TP|Media_SQE)) |
media_tbl[dev->if_port].media_bits, ioaddr + Wn4_Media);
}
/* Switch to the stats window, and clear all stats by reading. */
outw(StatsDisable, ioaddr + EL3_CMD);
EL3WINDOW(6);
for (i = 0; i < 10; i++)
inb(ioaddr + i);
inw(ioaddr + 10);
inw(ioaddr + 12);
/* New: On the Vortex we must also clear the BadSSD counter. */
EL3WINDOW(4);
inb(ioaddr + 12);
/* ..and on the Boomerang we enable the extra statistics bits. */
outw(0x0040, ioaddr + Wn4_NetDiag);
/* Switch to register set 7 for normal use. */
EL3WINDOW(7);
if (vp->full_bus_master_rx) { /* Boomerang bus master. */
vp->cur_rx = vp->dirty_rx = 0;
/* Initialize the RxEarly register as recommended. */
outw(SetRxThreshold + (1536>>2), ioaddr + EL3_CMD);
outl(0x0020, ioaddr + PktStatus);
outl(vp->rx_ring_dma, ioaddr + UpListPtr);
}
if (vp->full_bus_master_tx) { /* Boomerang bus master Tx. */
dev->hard_start_xmit = &boomerang_start_xmit;
vp->cur_tx = vp->dirty_tx = 0;
outb(PKT_BUF_SZ>>8, ioaddr + TxFreeThreshold); /* Room for a packet. */
/* Clear the Rx, Tx rings. */
for (i = 0; i < RX_RING_SIZE; i++) /* AKPM: this is done in vortex_open, too */
vp->rx_ring[i].status = 0;
for (i = 0; i < TX_RING_SIZE; i++)
vp->tx_skbuff[i] = 0;
outl(0, ioaddr + DownListPtr);
}
/* Set receiver mode: presumably accept b-case and phys addr only. */
set_rx_mode(dev);
outw(StatsEnable, ioaddr + EL3_CMD); /* Turn on statistics. */
netif_start_queue (dev);
outw(RxEnable, ioaddr + EL3_CMD); /* Enable the receiver. */
outw(TxEnable, ioaddr + EL3_CMD); /* Enable transmitter. */
/* Allow status bits to be seen. */
vp->status_enable = SetStatusEnb | HostError|IntReq|StatsFull|TxComplete|
(vp->full_bus_master_tx ? DownComplete : TxAvailable) |
(vp->full_bus_master_rx ? UpComplete : RxComplete) |
(vp->bus_master ? DMADone : 0);
vp->intr_enable = SetIntrEnb | IntLatch | TxAvailable |
(vp->full_bus_master_rx ? 0 : RxComplete) |
StatsFull | HostError | TxComplete | IntReq
| (vp->bus_master ? DMADone : 0) | UpComplete | DownComplete;
outw(vp->status_enable, ioaddr + EL3_CMD);
/* Ack all pending events, and set active indicator mask. */
outw(AckIntr | IntLatch | TxAvailable | RxEarly | IntReq,
ioaddr + EL3_CMD);
outw(vp->intr_enable, ioaddr + EL3_CMD);
if (vp->cb_fn_base) /* The PCMCIA people are idiots. */
writel(0x8000, vp->cb_fn_base + 4);
if (extra_reset)
{
/* AKPM: unjam the 3CCFE575CT */
wait_for_completion(dev, TxReset);
if (vp->full_bus_master_tx) {
outb(PKT_BUF_SZ>>8, ioaddr + TxFreeThreshold);
outw(DownUnstall, ioaddr + EL3_CMD);
}
outw(TxEnable, ioaddr + EL3_CMD);
}
}
static int
vortex_open(struct net_device *dev)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
int i;
int retval;
MOD_INC_USE_COUNT;
/* Use the now-standard shared IRQ implementation. */
if (request_irq(dev->irq, vp->full_bus_master_rx ? &boomerang_interrupt : &vortex_interrupt,
SA_SHIRQ, dev->name, dev)) {
retval = -EAGAIN;
goto out;
}
if (vp->full_bus_master_rx) { /* Boomerang bus master. */
if (vortex_debug > 2)
printk(KERN_DEBUG "%s: Filling in the Rx ring.\n", dev->name);
for (i = 0; i < RX_RING_SIZE; i++) {
struct sk_buff *skb;
vp->rx_ring[i].next = cpu_to_le32(vp->rx_ring_dma + sizeof(struct boom_rx_desc) * (i+1));
vp->rx_ring[i].status = 0; /* Clear complete bit. */
vp->rx_ring[i].length = cpu_to_le32(PKT_BUF_SZ | LAST_FRAG);
skb = dev_alloc_skb(PKT_BUF_SZ);
vp->rx_skbuff[i] = skb;
if (skb == NULL)
break; /* Bad news! */
skb->dev = dev; /* Mark as being used by this device. */
skb_reserve(skb, 2); /* Align IP on 16 byte boundaries */
vp->rx_ring[i].addr = cpu_to_le32(pci_map_single(vp->pdev, skb->tail, PKT_BUF_SZ, PCI_DMA_FROMDEVICE));
}
/* Wrap the ring. */
vp->rx_ring[i-1].next = cpu_to_le32(vp->rx_ring_dma);
}
vortex_up(dev);
vp->open = 1;
return 0;
out:
MOD_DEC_USE_COUNT;
if (vortex_debug > 1)
printk(KERN_ERR PFX "vortex_open() fails: returning %d\n", retval);
return retval;
}
static void vortex_timer(unsigned long data)
{
struct net_device *dev = (struct net_device *)data;
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr = dev->base_addr;
int next_tick = TX_EXPIRE;
int ok = 0;
int media_status, mii_status, old_window;
if (vortex_debug > 1)
printk(KERN_DEBUG "%s: Media selection timer tick happened, %s.\n",
dev->name, media_tbl[dev->if_port].name);
disable_irq(dev->irq);
old_window = inw(ioaddr + EL3_CMD) >> 13;
EL3WINDOW(4);
media_status = inw(ioaddr + Wn4_Media);
switch (dev->if_port) {
case XCVR_10baseT: case XCVR_100baseTx: case XCVR_100baseFx:
if (media_status & Media_LnkBeat) {
ok = 1;
netif_carrier_on(dev);
if (vortex_debug > 1)
printk(KERN_DEBUG "%s: Media %s has link beat, %x.\n",
dev->name, media_tbl[dev->if_port].name, media_status);
} else {
netif_carrier_off(dev);
if (vortex_debug > 1)
printk(KERN_DEBUG "%s: Media %s has no link beat, %x.\n",
dev->name, media_tbl[dev->if_port].name, media_status);
}
break;
case XCVR_MII: case XCVR_NWAY:
{
unsigned long flags;
spin_lock_irqsave(&vp->lock, flags); /* AKPM: protect mdio state */
if (media_status & Media_LnkBeat) {
netif_carrier_on(dev);
} else {
netif_carrier_off(dev);
}
mii_status = mdio_read(ioaddr, vp->phys[0], 1);
ok = 1;
if (vortex_debug > 1)
printk(KERN_DEBUG "%s: MII transceiver has status %4.4x.\n",
dev->name, mii_status);
if (mii_status & 0x0004) {
int mii_reg5 = mdio_read(ioaddr, vp->phys[0], 5);
if (! vp->force_fd && mii_reg5 != 0xffff) {
int duplex = (mii_reg5&0x0100) ||
(mii_reg5 & 0x01C0) == 0x0040;
if (vp->full_duplex != duplex) {
vp->full_duplex = duplex;
printk(KERN_INFO "%s: Setting %s-duplex based on MII "
"#%d link partner capability of %4.4x.\n",
dev->name, vp->full_duplex ? "full" : "half",
vp->phys[0], mii_reg5);
/* Set the full-duplex bit. */
EL3WINDOW(3); /* AKPM: this was missing from 2.3.99 3c59x.c! */
outb((vp->full_duplex ? 0x20 : 0) |
(dev->mtu > 1500 ? 0x40 : 0),
ioaddr + Wn3_MAC_Ctrl);
if (vortex_debug > 1)
printk(KERN_DEBUG "Setting duplex in Wn3_MAC_Ctrl\n");
/* AKPM: bug: should reset Tx and Rx after setting Duplex. Page 180 */
}
next_tick = TX_EXPIRE;
}
}
spin_unlock_irqrestore(&vp->lock, flags);
}
break;
default: /* Other media types handled by Tx timeouts. */
if (vortex_debug > 1)
printk(KERN_DEBUG "%s: Media %s has no indication, %x.\n",
dev->name, media_tbl[dev->if_port].name, media_status);
ok = 1;
}
if ( ! ok) {
union wn3_config config;
do {
dev->if_port = media_tbl[dev->if_port].next;
} while ( ! (vp->available_media & media_tbl[dev->if_port].mask));
if (dev->if_port == XCVR_Default) { /* Go back to default. */
dev->if_port = vp->default_media;
if (vortex_debug > 1)
printk(KERN_DEBUG "%s: Media selection failing, using default "
"%s port.\n",
dev->name, media_tbl[dev->if_port].name);
} else {
if (vortex_debug > 1)
printk(KERN_DEBUG "%s: Media selection failed, now trying "
"%s port.\n",
dev->name, media_tbl[dev->if_port].name);
next_tick = TX_EXPIRE;
// next_tick = media_tbl[dev->if_port].wait;
}
outw((media_status & ~(Media_10TP|Media_SQE)) |
media_tbl[dev->if_port].media_bits, ioaddr + Wn4_Media);
EL3WINDOW(3);
config.i = inl(ioaddr + Wn3_Config);
config.u.xcvr = dev->if_port;
outl(config.i, ioaddr + Wn3_Config);
outw(dev->if_port == XCVR_10base2 ? StartCoax : StopCoax,
ioaddr + EL3_CMD);
if (vortex_debug > 1)
printk(KERN_DEBUG "wrote 0x%08x to Wn3_Config\n", config.i);
/* AKPM: FIXME: Should reset Rx & Tx here. P60 of 3c90xc.pdf */
}
EL3WINDOW(old_window);
enable_irq(dev->irq);
if (vortex_debug > 1)
printk(KERN_DEBUG "%s: Media selection timer finished, %s.\n",
dev->name, media_tbl[dev->if_port].name);
vp->timer.expires = RUN_AT(next_tick);
add_timer(&vp->timer);
if (vp->deferred)
outw(FakeIntr, ioaddr + EL3_CMD);
return;
}
static void vortex_tx_timeout(struct net_device *dev)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr = dev->base_addr;
printk(KERN_ERR "%s: transmit timed out, tx_status %2.2x status %4.4x.\n",
dev->name, inb(ioaddr + TxStatus),
inw(ioaddr + EL3_STATUS));
/* Slight code bloat to be user friendly. */
if ((inb(ioaddr + TxStatus) & 0x88) == 0x88)
printk(KERN_ERR "%s: Transmitter encountered 16 collisions --"
" network cable problem?\n", dev->name);
if (inw(ioaddr + EL3_STATUS) & IntLatch) {
printk(KERN_ERR "%s: Interrupt posted but not delivered --"
" IRQ blocked by another device?\n", dev->name);
/* Bad idea here.. but we might as well handle a few events. */
{
/*
* AKPM: block interrupts because vortex_interrupt
* does a bare spin_lock()
*/
unsigned long flags;
local_irq_save(flags);
if (vp->full_bus_master_tx)
boomerang_interrupt(dev->irq, dev, 0);
else
vortex_interrupt(dev->irq, dev, 0);
local_irq_restore(flags);
}
}
if (vortex_debug > 1)
dump_tx_ring(dev);
wait_for_completion(dev, TxReset);
vp->stats.tx_errors++;
if (vp->full_bus_master_tx) {
if (vortex_debug > 0)
printk(KERN_DEBUG "%s: Resetting the Tx ring pointer.\n",
dev->name);
if (vp->cur_tx - vp->dirty_tx > 0 && inl(ioaddr + DownListPtr) == 0)
outl(vp->tx_ring_dma + (vp->dirty_tx % TX_RING_SIZE) * sizeof(struct boom_tx_desc),
ioaddr + DownListPtr);
if (vp->tx_full && (vp->cur_tx - vp->dirty_tx <= TX_RING_SIZE - 1)) {
vp->tx_full = 0;
netif_wake_queue (dev);
}
if (vp->tx_full)
netif_stop_queue (dev);
outb(PKT_BUF_SZ>>8, ioaddr + TxFreeThreshold);
outw(DownUnstall, ioaddr + EL3_CMD);
} else
vp->stats.tx_dropped++;
/* Issue Tx Enable */
outw(TxEnable, ioaddr + EL3_CMD);
dev->trans_start = jiffies;
/* Switch to register set 7 for normal use. */
EL3WINDOW(7);
}
/*
* Handle uncommon interrupt sources. This is a separate routine to minimize
* the cache impact.
*/
static void
vortex_error(struct net_device *dev, int status)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr = dev->base_addr;
int do_tx_reset = 0;
unsigned char tx_status = 0;
if (vortex_debug > 2) {
printk(KERN_DEBUG "%s: vortex_error(), status=0x%x\n", dev->name, status);
}
if (status & TxComplete) { /* Really "TxError" for us. */
tx_status = inb(ioaddr + TxStatus);
/* Presumably a tx-timeout. We must merely re-enable. */
if (vortex_debug > 2
|| (tx_status != 0x88 && vortex_debug > 0)) {
printk(KERN_DEBUG"%s: Transmit error, Tx status register %2.2x.\n",
dev->name, tx_status);
dump_tx_ring(dev);
}
if (tx_status & 0x14) vp->stats.tx_fifo_errors++;
if (tx_status & 0x38) vp->stats.tx_aborted_errors++;
outb(0, ioaddr + TxStatus);
if (tx_status & 0x38) /* AKPM: tx reset after 16 collisions, despite what the manual says */
do_tx_reset = 1;
else /* Merely re-enable the transmitter. */
outw(TxEnable, ioaddr + EL3_CMD);
}
if (status & RxEarly) { /* Rx early is unused. */
vortex_rx(dev);
outw(AckIntr | RxEarly, ioaddr + EL3_CMD);
}
if (status & StatsFull) { /* Empty statistics. */
static int DoneDidThat = 0;
if (vortex_debug > 4)
printk(KERN_DEBUG "%s: Updating stats.\n", dev->name);
update_stats(ioaddr, dev);
/* HACK: Disable statistics as an interrupt source. */
/* This occurs when we have the wrong media type! */
if (DoneDidThat == 0 &&
inw(ioaddr + EL3_STATUS) & StatsFull) {
printk(KERN_WARNING "%s: Updating statistics failed, disabling "
"stats as an interrupt source.\n", dev->name);
EL3WINDOW(5);
outw(SetIntrEnb | (inw(ioaddr + 10) & ~StatsFull), ioaddr + EL3_CMD);
vp->intr_enable &= ~StatsFull;
EL3WINDOW(7);
DoneDidThat++;
}
}
if (status & IntReq) { /* Restore all interrupt sources. */
outw(vp->status_enable, ioaddr + EL3_CMD);
outw(vp->intr_enable, ioaddr + EL3_CMD);
}
if (status & HostError) {
u16 fifo_diag;
EL3WINDOW(4);
fifo_diag = inw(ioaddr + Wn4_FIFODiag);
printk(KERN_ERR "%s: Host error, FIFO diagnostic register %4.4x.\n",
dev->name, fifo_diag);
/* Adapter failure requires Tx/Rx reset and reinit. */
if (vp->full_bus_master_tx) {
/* In this case, blow the card away */
vortex_down(dev);
wait_for_completion(dev, TotalReset | 0xff);
vortex_up(dev);
} else if (fifo_diag & 0x0400)
do_tx_reset = 1;
if (fifo_diag & 0x3000) {
wait_for_completion(dev, RxReset);
/* Set the Rx filter to the current state. */
set_rx_mode(dev);
outw(RxEnable, ioaddr + EL3_CMD); /* Re-enable the receiver. */
outw(AckIntr | HostError, ioaddr + EL3_CMD);
}
}
/*
* Black magic. If we're resetting the transmitter, remember the current downlist
* pointer and restore it afterwards. We can't usr cur_tx because that could
* lag the actual hardware index.
*/
if (do_tx_reset) {
if (vp->full_bus_master_tx) {
unsigned long old_down_list_ptr;
wait_for_completion(dev, DownStall);
old_down_list_ptr = inl(ioaddr + DownListPtr);
wait_for_completion(dev, TxReset);
outw(TxEnable, ioaddr + EL3_CMD);
/* Restart DMA if necessary */
outl(old_down_list_ptr, ioaddr + DownListPtr);
if (vortex_debug > 2)
printk(KERN_DEBUG "reset DMA to 0x%08x\n", inl(ioaddr + DownListPtr));
outw(DownUnstall, ioaddr + EL3_CMD);
/*
* Here we make a single attempt to prevent a timeout by
* restarting the timer if we think that the ISR has a good
* chance of unjamming things.
*/
if (vp->tx_reset_resume == 0 && vp->tx_full) {
vp->tx_reset_resume = 1;
dev->trans_start = jiffies;
}
} else {
wait_for_completion(dev, TxReset);
outw(TxEnable, ioaddr + EL3_CMD);
}
}
}
static int
vortex_start_xmit(struct sk_buff *skb, struct net_device *dev)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr = dev->base_addr;
/* Put out the doubleword header... */
outl(skb->len, ioaddr + TX_FIFO);
if (vp->bus_master) {
/* Set the bus-master controller to transfer the packet. */
int len = (skb->len + 3) & ~3;
outl( vp->tx_skb_dma = pci_map_single(vp->pdev, skb->data, len, PCI_DMA_TODEVICE),
ioaddr + Wn7_MasterAddr);
outw(len, ioaddr + Wn7_MasterLen);
vp->tx_skb = skb;
outw(StartDMADown, ioaddr + EL3_CMD);
/* netif_wake_queue() will be called at the DMADone interrupt. */
} else {
/* ... and the packet rounded to a doubleword. */
outsl(ioaddr + TX_FIFO, skb->data, (skb->len + 3) >> 2);
dev_kfree_skb (skb);
if (inw(ioaddr + TxFree) > 1536) {
netif_start_queue (dev); /* AKPM: redundant? */
} else {
/* Interrupt us when the FIFO has room for max-sized packet. */
netif_stop_queue(dev);
outw(SetTxThreshold + (1536>>2), ioaddr + EL3_CMD);
}
}
dev->trans_start = jiffies;
/* Clear the Tx status stack. */
{
int tx_status;
int i = 32;
while (--i > 0 && (tx_status = inb(ioaddr + TxStatus)) > 0) {
if (tx_status & 0x3C) { /* A Tx-disabling error occurred. */
if (vortex_debug > 2)
printk(KERN_DEBUG "%s: Tx error, status %2.2x.\n",
dev->name, tx_status);
if (tx_status & 0x04) vp->stats.tx_fifo_errors++;
if (tx_status & 0x38) vp->stats.tx_aborted_errors++;
if (tx_status & 0x30) {
wait_for_completion(dev, TxReset);
}
outw(TxEnable, ioaddr + EL3_CMD);
}
outb(0x00, ioaddr + TxStatus); /* Pop the status stack. */
}
}
return 0;
}
static int
boomerang_start_xmit(struct sk_buff *skb, struct net_device *dev)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr = dev->base_addr;
/* Calculate the next Tx descriptor entry. */
int entry = vp->cur_tx % TX_RING_SIZE;
struct boom_tx_desc *prev_entry = &vp->tx_ring[(vp->cur_tx-1) % TX_RING_SIZE];
unsigned long flags;
if (vortex_debug > 6)
printk(KERN_DEBUG "boomerang_start_xmit()\n");
if (vortex_debug > 3)
printk(KERN_DEBUG "%s: Trying to send a packet, Tx index %d.\n",
dev->name, vp->cur_tx);
if (vp->tx_full) {
if (vortex_debug > 0)
printk(KERN_WARNING "%s: Tx Ring full, refusing to send buffer.\n",
dev->name);
return 1;
}
vp->tx_skbuff[entry] = skb;
vp->tx_ring[entry].next = 0;
vp->tx_ring[entry].addr = cpu_to_le32(pci_map_single(vp->pdev, skb->data, skb->len, PCI_DMA_TODEVICE));
vp->tx_ring[entry].length = cpu_to_le32(skb->len | LAST_FRAG);
vp->tx_ring[entry].status = cpu_to_le32(skb->len | TxIntrUploaded);
spin_lock_irqsave(&vp->lock, flags);
/* Wait for the stall to complete. */
wait_for_completion(dev, DownStall);
prev_entry->next = cpu_to_le32(vp->tx_ring_dma + entry * sizeof(struct boom_tx_desc));
if (inl(ioaddr + DownListPtr) == 0) {
outl(vp->tx_ring_dma + entry * sizeof(struct boom_tx_desc), ioaddr + DownListPtr);
queued_packet++;
}
vp->cur_tx++;
if (vp->cur_tx - vp->dirty_tx > TX_RING_SIZE - 1) {
vp->tx_full = 1;
netif_stop_queue (dev);
} else { /* Clear previous interrupt enable. */
#if defined(tx_interrupt_mitigation)
prev_entry->status &= cpu_to_le32(~TxIntrUploaded);
#endif
/* netif_start_queue (dev); */ /* AKPM: redundant? */
}
outw(DownUnstall, ioaddr + EL3_CMD);
vp->tx_reset_resume = 0;
spin_unlock_irqrestore(&vp->lock, flags);
dev->trans_start = jiffies;
return 0;
}
/* The interrupt handler does all of the Rx thread work and cleans up
after the Tx thread. */
/*
* This is the ISR for the vortex series chips.
* full_bus_master_tx == 0 && full_bus_master_rx == 0
*/
static void vortex_interrupt(int irq, void *dev_id, struct pt_regs *regs)
{
struct net_device *dev = dev_id;
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr;
int status;
int work_done = max_interrupt_work;
spin_lock(&vp->lock);
ioaddr = dev->base_addr;
status = inw(ioaddr + EL3_STATUS);
if (vortex_debug > 6)
printk("vortex_interrupt(). status=0x%4x\n", status);
if (status & IntReq) {
status |= vp->deferred;
vp->deferred = 0;
}
if (status == 0xffff) /* AKPM: h/w no longer present (hotplug)? */
goto handler_exit;
if (vortex_debug > 4)
printk(KERN_DEBUG "%s: interrupt, status %4.4x, latency %d ticks.\n",
dev->name, status, inb(ioaddr + Timer));
do {
if (vortex_debug > 5)
printk(KERN_DEBUG "%s: In interrupt loop, status %4.4x.\n",
dev->name, status);
if (status & RxComplete)
vortex_rx(dev);
if (status & TxAvailable) {
if (vortex_debug > 5)
printk(KERN_DEBUG " TX room bit was handled.\n");
/* There's room in the FIFO for a full-sized packet. */
outw(AckIntr | TxAvailable, ioaddr + EL3_CMD);
netif_wake_queue (dev);
}
if (status & DMADone) {
if (inw(ioaddr + Wn7_MasterStatus) & 0x1000) {
outw(0x1000, ioaddr + Wn7_MasterStatus); /* Ack the event. */
pci_unmap_single(vp->pdev, vp->tx_skb_dma, (vp->tx_skb->len + 3) & ~3, PCI_DMA_TODEVICE);
dev_kfree_skb_irq(vp->tx_skb); /* Release the transferred buffer */
if (inw(ioaddr + TxFree) > 1536) {
/*
* AKPM: FIXME: I don't think we need this. If the queue was stopped due to
* insufficient FIFO room, the TxAvailable test will succeed and call
* netif_wake_queue()
*/
netif_wake_queue(dev);
} else { /* Interrupt when FIFO has room for max-sized packet. */
outw(SetTxThreshold + (1536>>2), ioaddr + EL3_CMD);
netif_stop_queue(dev); /* AKPM: This is new */
}
}
}
/* Check for all uncommon interrupts at once. */
if (status & (HostError | RxEarly | StatsFull | TxComplete | IntReq)) {
if (status == 0xffff)
break;
vortex_error(dev, status);
}
if (--work_done < 0) {
printk(KERN_WARNING "%s: Too much work in interrupt, status "
"%4.4x.\n", dev->name, status);
/* Disable all pending interrupts. */
do {
vp->deferred |= status;
outw(SetStatusEnb | (~vp->deferred & vp->status_enable),
ioaddr + EL3_CMD);
outw(AckIntr | (vp->deferred & 0x7ff), ioaddr + EL3_CMD);
} while ((status = inw(ioaddr + EL3_CMD)) & IntLatch);
/* The timer will reenable interrupts. */
mod_timer(&vp->timer, jiffies + 1*HZ);
break;
}
/* Acknowledge the IRQ. */
outw(AckIntr | IntReq | IntLatch, ioaddr + EL3_CMD);
if (vp->cb_fn_base) /* The PCMCIA people are idiots. */
writel(0x8000, vp->cb_fn_base + 4);
} while ((status = inw(ioaddr + EL3_STATUS)) & (IntLatch | RxComplete));
if (vortex_debug > 4)
printk(KERN_DEBUG "%s: exiting interrupt, status %4.4x.\n",
dev->name, status);
handler_exit:
spin_unlock(&vp->lock);
}
/*
* This is the ISR for the boomerang series chips.
* full_bus_master_tx == 1 && full_bus_master_rx == 1
*/
static void boomerang_interrupt(int irq, void *dev_id, struct pt_regs *regs)
{
struct net_device *dev = dev_id;
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr;
int status;
int work_done = max_interrupt_work;
spin_lock(&vp->lock);
ioaddr = dev->base_addr;
status = inw(ioaddr + EL3_STATUS);
if (vortex_debug > 6)
printk(KERN_DEBUG "boomerang_interrupt. status=0x%4x\n", status);
if (status == 0xffff) { /* AKPM: h/w no longer present (hotplug)? */
if (vortex_debug > 1)
printk(KERN_DEBUG "boomerang_interrupt(1): status = 0xffff\n");
goto handler_exit;
}
if (status & IntReq) {
status |= vp->deferred;
vp->deferred = 0;
}
if (vortex_debug > 4)
printk(KERN_DEBUG "%s: interrupt, status %4.4x, latency %d ticks.\n",
dev->name, status, inb(ioaddr + Timer));
do {
if (vortex_debug > 5)
printk(KERN_DEBUG "%s: In interrupt loop, status %4.4x.\n",
dev->name, status);
if (status & UpComplete) {
outw(AckIntr | UpComplete, ioaddr + EL3_CMD);
if (vortex_debug > 5)
printk(KERN_DEBUG "boomerang_interrupt->boomerang_rx\n");
boomerang_rx(dev);
}
if (status & DownComplete) {
unsigned int dirty_tx = vp->dirty_tx;
outw(AckIntr | DownComplete, ioaddr + EL3_CMD);
while (vp->cur_tx - dirty_tx > 0) {
int entry = dirty_tx % TX_RING_SIZE;
if (inl(ioaddr + DownListPtr) ==
vp->tx_ring_dma + entry * sizeof(struct boom_tx_desc))
break; /* It still hasn't been processed. */
if (vp->tx_skbuff[entry]) {
struct sk_buff *skb = vp->tx_skbuff[entry];
pci_unmap_single(vp->pdev,
le32_to_cpu(vp->tx_ring[entry].addr), skb->len, PCI_DMA_TODEVICE);
dev_kfree_skb_irq(skb);
vp->tx_skbuff[entry] = 0;
} else {
printk(KERN_DEBUG "boomerang_interrupt: no skb!\n");
}
/* vp->stats.tx_packets++; Counted below. */
dirty_tx++;
}
vp->dirty_tx = dirty_tx;
if (vp->tx_full && (vp->cur_tx - dirty_tx <= TX_RING_SIZE - 1)) {
if (vortex_debug > 6)
printk(KERN_DEBUG "boomerang_interrupt: clearing tx_full\n");
vp->tx_full = 0;
netif_wake_queue (dev);
}
}
if (vp->tx_full)
netif_stop_queue (dev);
/* Check for all uncommon interrupts at once. */
if (status & (HostError | RxEarly | StatsFull | TxComplete | IntReq))
vortex_error(dev, status);
if (--work_done < 0) {
printk(KERN_WARNING "%s: Too much work in interrupt, status "
"%4.4x.\n", dev->name, status);
/* Disable all pending interrupts. */
do {
vp->deferred |= status;
outw(SetStatusEnb | (~vp->deferred & vp->status_enable),
ioaddr + EL3_CMD);
outw(AckIntr | (vp->deferred & 0x7ff), ioaddr + EL3_CMD);
} while ((status = inw(ioaddr + EL3_CMD)) & IntLatch);
/* The timer will reenable interrupts. */
mod_timer(&vp->timer, jiffies + 1*HZ);
break;
}
/* Acknowledge the IRQ. */
outw(AckIntr | IntReq | IntLatch, ioaddr + EL3_CMD);
if (vp->cb_fn_base) /* The PCMCIA people are idiots. */
writel(0x8000, vp->cb_fn_base + 4);
} while ((status = inw(ioaddr + EL3_STATUS)) & (IntLatch | RxComplete));
if (vortex_debug > 4)
printk(KERN_DEBUG "%s: exiting interrupt, status %4.4x.\n",
dev->name, status);
handler_exit:
spin_unlock(&vp->lock);
}
static int vortex_rx(struct net_device *dev)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr = dev->base_addr;
int i;
short rx_status;
if (vortex_debug > 5)
printk(KERN_DEBUG "vortex_rx(): status %4.4x, rx_status %4.4x.\n",
inw(ioaddr+EL3_STATUS), inw(ioaddr+RxStatus));
while ((rx_status = inw(ioaddr + RxStatus)) > 0) {
if (rx_status & 0x4000) { /* Error, update stats. */
unsigned char rx_error = inb(ioaddr + RxErrors);
if (vortex_debug > 2)
printk(KERN_DEBUG " Rx error: status %2.2x.\n", rx_error);
vp->stats.rx_errors++;
if (rx_error & 0x01) vp->stats.rx_over_errors++;
if (rx_error & 0x02) vp->stats.rx_length_errors++;
if (rx_error & 0x04) vp->stats.rx_frame_errors++;
if (rx_error & 0x08) vp->stats.rx_crc_errors++;
if (rx_error & 0x10) vp->stats.rx_length_errors++;
} else {
/* The packet length: up to 4.5K!. */
int pkt_len = rx_status & 0x1fff;
struct sk_buff *skb;
skb = dev_alloc_skb(pkt_len + 5);
if (vortex_debug > 4)
printk(KERN_DEBUG "Receiving packet size %d status %4.4x.\n",
pkt_len, rx_status);
if (skb != NULL) {
skb->dev = dev;
skb_reserve(skb, 2); /* Align IP on 16 byte boundaries */
/* 'skb_put()' points to the start of sk_buff data area. */
if (vp->bus_master &&
! (inw(ioaddr + Wn7_MasterStatus) & 0x8000)) {
dma_addr_t dma = pci_map_single(vp->pdev, skb_put(skb, pkt_len),
pkt_len, PCI_DMA_FROMDEVICE);
outl(dma, ioaddr + Wn7_MasterAddr);
outw((skb->len + 3) & ~3, ioaddr + Wn7_MasterLen);
outw(StartDMAUp, ioaddr + EL3_CMD);
while (inw(ioaddr + Wn7_MasterStatus) & 0x8000)
;
pci_unmap_single(vp->pdev, dma, pkt_len, PCI_DMA_FROMDEVICE);
} else {
insl(ioaddr + RX_FIFO, skb_put(skb, pkt_len),
(pkt_len + 3) >> 2);
}
outw(RxDiscard, ioaddr + EL3_CMD); /* Pop top Rx packet. */
skb->protocol = eth_type_trans(skb, dev);
netif_rx(skb);
dev->last_rx = jiffies;
vp->stats.rx_packets++;
/* Wait a limited time to go to next packet. */
for (i = 200; i >= 0; i--)
if ( ! (inw(ioaddr + EL3_STATUS) & CmdInProgress))
break;
continue;
} else if (vortex_debug > 0)
printk(KERN_NOTICE "%s: No memory to allocate a sk_buff of "
"size %d.\n", dev->name, pkt_len);
}
vp->stats.rx_dropped++;
wait_for_completion(dev, RxDiscard);
}
return 0;
}
static int
boomerang_rx(struct net_device *dev)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
int entry = vp->cur_rx % RX_RING_SIZE;
long ioaddr = dev->base_addr;
int rx_status;
int rx_work_limit = vp->dirty_rx + RX_RING_SIZE - vp->cur_rx;
if (vortex_debug > 5)
printk(KERN_DEBUG "boomerang_rx(): status %4.4x\n", inw(ioaddr+EL3_STATUS));
while ((rx_status = le32_to_cpu(vp->rx_ring[entry].status)) & RxDComplete){
if (--rx_work_limit < 0)
break;
if (rx_status & RxDError) { /* Error, update stats. */
unsigned char rx_error = rx_status >> 16;
if (vortex_debug > 2)
printk(KERN_DEBUG " Rx error: status %2.2x.\n", rx_error);
vp->stats.rx_errors++;
if (rx_error & 0x01) vp->stats.rx_over_errors++;
if (rx_error & 0x02) vp->stats.rx_length_errors++;
if (rx_error & 0x04) vp->stats.rx_frame_errors++;
if (rx_error & 0x08) vp->stats.rx_crc_errors++;
if (rx_error & 0x10) vp->stats.rx_length_errors++;
} else {
/* The packet length: up to 4.5K!. */
int pkt_len = rx_status & 0x1fff;
struct sk_buff *skb;
dma_addr_t dma = le32_to_cpu(vp->rx_ring[entry].addr);
vp->stats.rx_bytes += pkt_len;
if (vortex_debug > 4)
printk(KERN_DEBUG "Receiving packet size %d status %4.4x.\n",
pkt_len, rx_status);
/* Check if the packet is long enough to just accept without
copying to a properly sized skbuff. */
if (pkt_len < rx_copybreak
&& (skb = dev_alloc_skb(pkt_len + 2)) != 0) {
skb->dev = dev;
skb_reserve(skb, 2); /* Align IP on 16 byte boundaries */
pci_dma_sync_single(vp->pdev, dma, PKT_BUF_SZ, PCI_DMA_FROMDEVICE);
/* 'skb_put()' points to the start of sk_buff data area. */
memcpy(skb_put(skb, pkt_len),
vp->rx_skbuff[entry]->tail,
pkt_len);
rx_copy++;
} else {
/* Pass up the skbuff already on the Rx ring. */
skb = vp->rx_skbuff[entry];
vp->rx_skbuff[entry] = NULL;
skb_put(skb, pkt_len);
pci_unmap_single(vp->pdev, dma, PKT_BUF_SZ, PCI_DMA_FROMDEVICE);
rx_nocopy++;
}
skb->protocol = eth_type_trans(skb, dev);
{ /* Use hardware checksum info. */
int csum_bits = rx_status & 0xee000000;
if (csum_bits &&
(csum_bits == (IPChksumValid | TCPChksumValid) ||
csum_bits == (IPChksumValid | UDPChksumValid))) {
skb->ip_summed = CHECKSUM_UNNECESSARY;
rx_csumhits++;
}
}
netif_rx(skb);
dev->last_rx = jiffies;
vp->stats.rx_packets++;
}
entry = (++vp->cur_rx) % RX_RING_SIZE;
}
/* Refill the Rx ring buffers. */
for (; vp->dirty_rx < vp->cur_rx; vp->dirty_rx++) {
struct sk_buff *skb;
entry = vp->dirty_rx % RX_RING_SIZE;
if (vp->rx_skbuff[entry] == NULL) {
skb = dev_alloc_skb(PKT_BUF_SZ);
if (skb == NULL)
break; /* Bad news! */
skb->dev = dev; /* Mark as being used by this device. */
skb_reserve(skb, 2); /* Align IP on 16 byte boundaries */
vp->rx_ring[entry].addr = cpu_to_le32(pci_map_single(vp->pdev, skb->tail, PKT_BUF_SZ, PCI_DMA_FROMDEVICE));
vp->rx_skbuff[entry] = skb;
}
vp->rx_ring[entry].status = 0; /* Clear complete bit. */
outw(UpUnstall, ioaddr + EL3_CMD);
}
return 0;
}
static void
vortex_down(struct net_device *dev)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr = dev->base_addr;
netif_stop_queue (dev);
del_timer(&vp->timer);
/* Turn off statistics ASAP. We update vp->stats below. */
outw(StatsDisable, ioaddr + EL3_CMD);
/* Disable the receiver and transmitter. */
outw(RxDisable, ioaddr + EL3_CMD);
outw(TxDisable, ioaddr + EL3_CMD);
if (dev->if_port == XCVR_10base2)
/* Turn off thinnet power. Green! */
outw(StopCoax, ioaddr + EL3_CMD);
outw(SetIntrEnb | 0x0000, ioaddr + EL3_CMD);
update_stats(ioaddr, dev);
if (vp->full_bus_master_rx)
outl(0, ioaddr + UpListPtr);
if (vp->full_bus_master_tx)
outl(0, ioaddr + DownListPtr);
if (vp->capabilities & CapPwrMgmt)
acpi_set_WOL(dev);
}
static int
vortex_close(struct net_device *dev)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr = dev->base_addr;
int i;
if (netif_device_present(dev))
vortex_down(dev);
if (vortex_debug > 1) {
printk(KERN_DEBUG"%s: vortex_close() status %4.4x, Tx status %2.2x.\n",
dev->name, inw(ioaddr + EL3_STATUS), inb(ioaddr + TxStatus));
printk(KERN_DEBUG "%s: vortex close stats: rx_nocopy %d rx_copy %d"
" tx_queued %d Rx pre-checksummed %d.\n",
dev->name, rx_nocopy, rx_copy, queued_packet, rx_csumhits);
}
free_irq(dev->irq, dev);
if (vp->full_bus_master_rx) { /* Free Boomerang bus master Rx buffers. */
for (i = 0; i < RX_RING_SIZE; i++)
if (vp->rx_skbuff[i]) {
pci_unmap_single( vp->pdev, le32_to_cpu(vp->rx_ring[i].addr),
PKT_BUF_SZ, PCI_DMA_FROMDEVICE);
dev_kfree_skb(vp->rx_skbuff[i]);
vp->rx_skbuff[i] = 0;
}
}
if (vp->full_bus_master_tx) { /* Free Boomerang bus master Tx buffers. */
for (i = 0; i < TX_RING_SIZE; i++)
if (vp->tx_skbuff[i]) {
struct sk_buff *skb = vp->tx_skbuff[i];
pci_unmap_single(vp->pdev, le32_to_cpu(vp->tx_ring[i].addr), skb->len, PCI_DMA_TODEVICE);
dev_kfree_skb(skb);
vp->tx_skbuff[i] = 0;
}
}
MOD_DEC_USE_COUNT;
vp->open = 0;
return 0;
}
static void
dump_tx_ring(struct net_device *dev)
{
if (vortex_debug > 0) {
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr = dev->base_addr;
if (vp->full_bus_master_tx) {
int i;
int stalled = inl(ioaddr + PktStatus) & 0x04; /* Possible racy. But it's only debug stuff */
wait_for_completion(dev, DownStall);
printk(KERN_DEBUG " Flags; bus-master %d, full %d; dirty %d(%d) "
"current %d(%d).\n",
vp->full_bus_master_tx, vp->tx_full,
vp->dirty_tx, vp->dirty_tx % TX_RING_SIZE,
vp->cur_tx, vp->cur_tx % TX_RING_SIZE);
printk(KERN_DEBUG " Transmit list %8.8x vs. %p.\n",
inl(ioaddr + DownListPtr),
&vp->tx_ring[vp->dirty_tx % TX_RING_SIZE]);
for (i = 0; i < TX_RING_SIZE; i++) {
printk(KERN_DEBUG " %d: @%p length %8.8x status %8.8x\n", i,
&vp->tx_ring[i],
le32_to_cpu(vp->tx_ring[i].length),
le32_to_cpu(vp->tx_ring[i].status));
}
if (!stalled)
outw(DownUnstall, ioaddr + EL3_CMD);
}
}
}
static struct net_device_stats *vortex_get_stats(struct net_device *dev)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
unsigned long flags;
if (netif_device_present(dev)) { /* AKPM: Used to be netif_running */
spin_lock_irqsave (&vp->lock, flags);
update_stats(dev->base_addr, dev);
spin_unlock_irqrestore (&vp->lock, flags);
}
return &vp->stats;
}
/* Update statistics.
Unlike with the EL3 we need not worry about interrupts changing
the window setting from underneath us, but we must still guard
against a race condition with a StatsUpdate interrupt updating the
table. This is done by checking that the ASM (!) code generated uses
atomic updates with '+='.
*/
static void update_stats(long ioaddr, struct net_device *dev)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
int old_window = inw(ioaddr + EL3_CMD);
if (old_window == 0xffff) /* Chip suspended or ejected. */
return;
/* Unlike the 3c5x9 we need not turn off stats updates while reading. */
/* Switch to the stats window, and read everything. */
EL3WINDOW(6);
vp->stats.tx_carrier_errors += inb(ioaddr + 0);
vp->stats.tx_heartbeat_errors += inb(ioaddr + 1);
/* Multiple collisions. */ inb(ioaddr + 2);
vp->stats.collisions += inb(ioaddr + 3);
vp->stats.tx_window_errors += inb(ioaddr + 4);
vp->stats.rx_fifo_errors += inb(ioaddr + 5);
vp->stats.tx_packets += inb(ioaddr + 6);
vp->stats.tx_packets += (inb(ioaddr + 9)&0x30) << 4;
/* Rx packets */ inb(ioaddr + 7); /* Must read to clear */
/* Tx deferrals */ inb(ioaddr + 8);
/* Don't bother with register 9, an extension of registers 6&7.
If we do use the 6&7 values the atomic update assumption above
is invalid. */
vp->stats.rx_bytes += inw(ioaddr + 10);
vp->stats.tx_bytes += inw(ioaddr + 12);
/* New: On the Vortex we must also clear the BadSSD counter. */
EL3WINDOW(4);
inb(ioaddr + 12);
{
u8 up = inb(ioaddr + 13);
vp->stats.rx_bytes += (up & 0x0f) << 16;
vp->stats.tx_bytes += (up & 0xf0) << 12;
}
/* We change back to window 7 (not 1) with the Vortex. */
/* AKPM: the previous comment is obsolete - we switch back to the old window */
EL3WINDOW(old_window >> 13);
return;
}
static int vortex_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr = dev->base_addr;
u16 *data = (u16 *)&rq->ifr_data;
int phy = vp->phys[0] & 0x1f;
int retval;
unsigned long flags;
spin_lock_irqsave(&vp->lock, flags);
switch(cmd) {
case SIOCDEVPRIVATE: /* Get the address of the PHY in use. */
data[0] = phy;
case SIOCDEVPRIVATE+1: /* Read the specified MII register. */
EL3WINDOW(4);
data[3] = mdio_read(ioaddr, data[0] & 0x1f, data[1] & 0x1f);
retval = 0;
break;
case SIOCDEVPRIVATE+2: /* Write the specified MII register */
if (!capable(CAP_NET_ADMIN)) {
retval = -EPERM;
} else {
EL3WINDOW(4);
mdio_write(ioaddr, data[0] & 0x1f, data[1] & 0x1f, data[2]);
retval = 0;
}
break;
default:
retval = -EOPNOTSUPP;
break;
}
spin_unlock_irqrestore(&vp->lock, flags);
return retval;
}
/* Pre-Cyclone chips have no documented multicast filter, so the only
multicast setting is to receive all multicast frames. At least
the chip has a very clean way to set the mode, unlike many others. */
static void set_rx_mode(struct net_device *dev)
{
long ioaddr = dev->base_addr;
int new_mode;
if (dev->flags & IFF_PROMISC) {
if (vortex_debug > 0)
printk(KERN_NOTICE "%s: Setting promiscuous mode.\n", dev->name);
new_mode = SetRxFilter|RxStation|RxMulticast|RxBroadcast|RxProm;
} else if ((dev->mc_list) || (dev->flags & IFF_ALLMULTI)) {
new_mode = SetRxFilter|RxStation|RxMulticast|RxBroadcast;
} else
new_mode = SetRxFilter | RxStation | RxBroadcast;
outw(new_mode, ioaddr + EL3_CMD);
}
/* MII transceiver control section.
Read and write the MII registers using software-generated serial
MDIO protocol. See the MII specifications or DP83840A data sheet
for details. */
/* The maximum data clock rate is 2.5 Mhz. The minimum timing is usually
met by back-to-back PCI I/O cycles, but we insert a delay to avoid
"overclocking" issues. */
#define mdio_delay() inl(mdio_addr)
#define MDIO_SHIFT_CLK 0x01
#define MDIO_DIR_WRITE 0x04
#define MDIO_DATA_WRITE0 (0x00 | MDIO_DIR_WRITE)
#define MDIO_DATA_WRITE1 (0x02 | MDIO_DIR_WRITE)
#define MDIO_DATA_READ 0x02
#define MDIO_ENB_IN 0x00
/* Generate the preamble required for initial synchronization and
a few older transceivers. */
static void mdio_sync(long ioaddr, int bits)
{
long mdio_addr = ioaddr + Wn4_PhysicalMgmt;
/* Establish sync by sending at least 32 logic ones. */
while (-- bits >= 0) {
outw(MDIO_DATA_WRITE1, mdio_addr);
mdio_delay();
outw(MDIO_DATA_WRITE1 | MDIO_SHIFT_CLK, mdio_addr);
mdio_delay();
}
}
static int mdio_read(long ioaddr, int phy_id, int location)
{
int i;
int read_cmd = (0xf6 << 10) | (phy_id << 5) | location;
unsigned int retval = 0;
long mdio_addr = ioaddr + Wn4_PhysicalMgmt;
if (mii_preamble_required)
mdio_sync(ioaddr, 32);
/* Shift the read command bits out. */
for (i = 14; i >= 0; i--) {
int dataval = (read_cmd&(1<<i)) ? MDIO_DATA_WRITE1 : MDIO_DATA_WRITE0;
outw(dataval, mdio_addr);
mdio_delay();
outw(dataval | MDIO_SHIFT_CLK, mdio_addr);
mdio_delay();
}
/* Read the two transition, 16 data, and wire-idle bits. */
for (i = 19; i > 0; i--) {
outw(MDIO_ENB_IN, mdio_addr);
mdio_delay();
retval = (retval << 1) | ((inw(mdio_addr) & MDIO_DATA_READ) ? 1 : 0);
outw(MDIO_ENB_IN | MDIO_SHIFT_CLK, mdio_addr);
mdio_delay();
}
return retval & 0x20000 ? 0xffff : retval>>1 & 0xffff;
}
static void mdio_write(long ioaddr, int phy_id, int location, int value)
{
int write_cmd = 0x50020000 | (phy_id << 23) | (location << 18) | value;
long mdio_addr = ioaddr + Wn4_PhysicalMgmt;
int i;
if (mii_preamble_required)
mdio_sync(ioaddr, 32);
/* Shift the command bits out. */
for (i = 31; i >= 0; i--) {
int dataval = (write_cmd&(1<<i)) ? MDIO_DATA_WRITE1 : MDIO_DATA_WRITE0;
outw(dataval, mdio_addr);
mdio_delay();
outw(dataval | MDIO_SHIFT_CLK, mdio_addr);
mdio_delay();
}
/* Leave the interface idle. */
for (i = 1; i >= 0; i--) {
outw(MDIO_ENB_IN, mdio_addr);
mdio_delay();
outw(MDIO_ENB_IN | MDIO_SHIFT_CLK, mdio_addr);
mdio_delay();
}
return;
}
\f
/* ACPI: Advanced Configuration and Power Interface. */
/* Set Wake-On-LAN mode and put the board into D3 (power-down) state. */
static void acpi_set_WOL(struct net_device *dev)
{
struct vortex_private *vp = (struct vortex_private *)dev->priv;
long ioaddr = dev->base_addr;
/* AKPM: This kills the 905 */
if (vortex_debug > 0) {
printk(KERN_INFO PFX "Wake-on-LAN functions disabled\n");
}
return;
/* Power up on: 1==Downloaded Filter, 2==Magic Packets, 4==Link Status. */
EL3WINDOW(7);
outw(2, ioaddr + 0x0c);
/* The RxFilter must accept the WOL frames. */
outw(SetRxFilter|RxStation|RxMulticast|RxBroadcast, ioaddr + EL3_CMD);
outw(RxEnable, ioaddr + EL3_CMD);
/* Change the power state to D3; RxEnable doesn't take effect. */
pci_write_config_word(vp->pdev, 0xe0, 0x8103);
}
static void __devexit vortex_remove_one (struct pci_dev *pdev)
{
struct net_device *dev = pdev->driver_data;
struct vortex_private *vp;
if (!dev) {
printk("vortex_remove_one called for EISA device!\n");
BUG();
}
vp = (void *)(dev->priv);
/* No need to check MOD_IN_USE, as sys_delete_module() checks. */
/* AKPM: FIXME: we should have
* if (vp->cb_fn_base) iounmap(vp->cb_fn_base);
* here
*/
unregister_netdev(dev);
outw(TotalReset, dev->base_addr + EL3_CMD);
release_region(dev->base_addr, vp->io_size);
kfree(dev);
}
static struct pci_driver vortex_driver = {
name: "3c575_cb",
probe: vortex_init_one,
remove: vortex_remove_one,
suspend: vortex_suspend,
resume: vortex_resume,
id_table: vortex_pci_tbl,
};
static int vortex_have_pci = 0;
static int vortex_have_eisa = 0;
static int __init vortex_init (void)
{
int rc;
rc = pci_module_init (&vortex_driver);
if (rc < 0)
goto out;
if (rc >= 0) /* AKPM: had "> 0" */
vortex_have_pci = 1;
rc = vortex_eisa_init ();
if (rc < 0)
goto out;
if (rc > 0)
vortex_have_eisa = 1;
out:
return rc;
}
static void __exit vortex_eisa_cleanup (void)
{
struct net_device *dev, *tmp;
struct vortex_private *vp;
long ioaddr;
dev = root_vortex_eisa_dev;
while (dev) {
vp = dev->priv;
ioaddr = dev->base_addr;
unregister_netdev (dev);
outw (TotalReset, ioaddr + EL3_CMD);
release_region (ioaddr, VORTEX_TOTAL_SIZE);
tmp = dev;
dev = vp->next_module;
kfree (tmp);
}
}
static void __exit vortex_cleanup (void)
{
if (vortex_have_pci)
pci_unregister_driver (&vortex_driver);
if (vortex_have_eisa)
vortex_eisa_cleanup ();
}
module_init(vortex_init);
module_exit(vortex_cleanup);
\f
/*
* Local variables:
* compile-command: "gcc -DMODULE -D__KERNEL__ -Wall -Wstrict-prototypes -O6 -c 3c59x.c `[ -f /usr/include/linux/modversions.h ] && echo -DMODVERSIONS`"
* cardbus-compile-command: "gcc -DCONFIG_HOTPLUG -DMODULE -D__KERNEL__ -Wall -Wstrict-prototypes -O6 -c 3c59x.c -o 3c575_cb.o -I/usr/src/linux/pcmcia-cs-3.0.9/include/"
* c-indent-level: 4
* c-basic-offset: 4
* tab-width: 4
* End:
*/
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [Bonding-devel] Re: Bonding driver unreliable under high CPUload
2002-09-17 20:11 ` Jeff Garzik
2002-09-17 20:24 ` Andrew Morton
@ 2002-09-17 20:30 ` Andrew Morton
1 sibling, 0 replies; 13+ messages in thread
From: Andrew Morton @ 2002-09-17 20:30 UTC (permalink / raw)
To: Jeff Garzik; +Cc: Cureington, Tony, Pascal Brisset, bonding-devel, netdev
(bonding-devel didn't like the size of the attachment.
It's at http://www.zip.com.au/~akpm/3c59x.c-netif)
Jeff Garzik wrote:
>
> Andrew Morton wrote:
> > Jeff Garzik wrote:
> >
> >>...
> >>Also, a further question: do you have access to the slave struct
> >>net_device? If so, just test netif_carrier_ok(slave_dev) and avoid all
> >>that ioctl calling if it returns non-zero.
> >
> >
> > Make that "avoid all that ioctl calling from interrupt context", which
> > is a bug. Of the box-killing variety ;)
>
> Indeed. /me looks at the bond_check_dev_link callers more closely and
> shudders.
>
> That wants fixing...
>
> Note that netif_carrier_ok() can indeed be checked in interrupt context.
> And if someone wants to send me patches converting more drivers to use
> netif_carrier_{on,off}, I would be very happy :)
That would be best.
I'm so slack. I received the below two years ago; Nelson has
added netif_carrier_foo support to 3c59x.c. As Jeff says: patches
solicited.
-------- Original Message --------
Subject: Re: netlink
Date: Wed, 14 Jun 2000 11:19:33 +0800 (SST)
From: Nelson <tanginhw@comp.nus.edu.sg>
To: Andrew Morton <andrewm@uow.edu.au>
References: <394610FF.4052CEAA@uow.edu.au>
hi Andrew,
The modified 3c59x.c is attached.
for your easy reference, these lines in the attached driver
was modified:
69-72: notes on what are added
121-123: defined the time to expiry of the vertex_timer
as TX_EXPIRE. i have set the value as 1*HZ.
the original value should be 60*HZ. you may
set to any value you feel is good ;p you r rite
abt the 1 sec part since reading the MII management
registers is time consuming.
1122: used TX_EXPIRE instead
1335, 1399, 1428:
set next_tick to TX_EXPIRE
1351: added calls to netif_carrier_on()
1356: added calls to netif_carrier_off()
1367-1371: added checking of link state
do let me know if there are any discrepencies.
thanx.
^ permalink raw reply [flat|nested] 13+ messages in thread
end of thread, other threads:[~2002-09-17 20:30 UTC | newest]
Thread overview: 13+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2002-09-17 19:28 [Bonding-devel] Re: Bonding driver unreliable under high CPU load Cureington, Tony
2002-09-17 19:45 ` Jeff Garzik
2002-09-17 19:53 ` [Bonding-devel] Re: Bonding driver unreliable under high CPUload Andrew Morton
2002-09-17 19:58 ` Chad N. Tindel
2002-09-17 20:07 ` Jeff Garzik
2002-09-17 20:11 ` Chad N. Tindel
2002-09-17 20:23 ` Jeff Garzik
2002-09-17 20:11 ` Jeff Garzik
2002-09-17 20:24 ` Andrew Morton
2002-09-17 20:30 ` Andrew Morton
-- strict thread matches above, loose matches on Subject: below --
2002-09-17 19:46 [Bonding-devel] Re: Bonding driver unreliable under high CPU load Jay Vosburgh
2002-09-17 20:01 Jay Vosburgh
2002-09-17 20:15 ` Jeff Garzik
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).